Classifications

• • G—PHYSICS
  • G09—EDUCATION; CRYPTOGRAPHY; DISPLAY; ADVERTISING; SEALS
  • G09B—EDUCATIONAL OR DEMONSTRATION APPLIANCES; APPLIANCES FOR TEACHING, OR COMMUNICATING WITH, THE BLIND, DEAF OR MUTE; MODELS; PLANETARIA; GLOBES; MAPS; DIAGRAMS
  • G09B29/00—Maps; Plans; Charts; Diagrams, e.g. route diagram
  • G09B29/10—Map spot or coordinate position indicators; Map reading aids
  • G09B29/106—Map spot or coordinate position indicators; Map reading aids using electronic means
• • G—PHYSICS
  • G01—MEASURING; TESTING
  • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
  • G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
  • G01C21/26—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 specially adapted for navigation in a road network
  • G01C21/34—Route searching; Route guidance
• • G—PHYSICS
  • G01—MEASURING; TESTING
  • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
  • G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
  • G01C21/26—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 specially adapted for navigation in a road network
  • G01C21/34—Route searching; Route guidance
  • G01C21/3446—Details of route searching algorithms, e.g. Dijkstra, A*, arc-flags or using precalculated routes
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
  • G06Q10/00—Administration; Management
  • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
  • G06Q10/047—Optimisation of routes or paths, e.g. travelling salesman problem
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
  • G06Q30/00—Commerce
  • G06Q30/02—Marketing; Price estimation or determination; Fundraising
  • G06Q30/0283—Price estimation or determination
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L45/00—Routing or path finding of packets in data switching networks
  • H04L45/12—Shortest path evaluation
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L45/00—Routing or path finding of packets in data switching networks
  • H04L45/34—Source routing

Definitions

• the present invention relates to a method of and an apparatus for generating a plurality of diverse routes in a weighted directed graph.
• reverse routes means routes which share the same links over less than a predetermined proportion of their lengths, typically over less than 85% of their lengths.
• An edge connects two vertices u and v; v is said to be adjacent u.
• each edge has a sense of direction from u to v and is written as an ordered pair ⁇ u,v> or u->v.
• an edge has no sense of direction and is written as an unordered pair ⁇ u,v ⁇ or u ⁇ ->v.
• An undirected graph can be represented by a directed graph if every undirected edge ⁇ u,v ⁇ is represented by two directed edges ⁇ u,v> and ⁇ v,u>.
• Both directed and undirected graphs may be weighted.
• a weight is attached to each edge. This may be used, for example, to represent the distance between two cities, the driving time, the cost of the journey, the resistance of an electrical path or some other quantity associated with the edge.
• the weight is sometimes called the length or cost of the edge, particularly when the graph represents a map of some kind.
• the weight or length of a path or a cycle is the sum of the weights or lengths of its component edges.
• the methods described in this application are generally applicable to any domain where the costs can be described by a weighted directed graph, and where there are no cycles of negative cost in the graph.
• Road route planners are designed to find the optimum route from source A to destination B. They define this as the single route with the minimum cost, where the cost function is a simple weighted sum of times, junction delays, distances, financial costs, type of road of the links that make up the route.
• shortest-path The methods are often called "shortest-path” where this is understood to mean shortest in the sum of the costs, whatever they represent, not necessarily in metres.
• shortest-path For a given cost function and road graph, there exists a lowest cost going from A to B, and this be found by using well known algorithms such as Dijkstra's algorithm. In a degenerate case, there may be a plurality of routes which share this globally lowest cost but the algorithms will find just one of them. Dijkstra's algorithm actually finds the optimum route from A to all nodes, but variations on it are designed to terminate early and to explore preferentially in the direction of B. The main constraint on the graph is that there should be no cycles with a negative overall cost, which is simple enough to satisfy for a road network where all the costs are positive.
• EP 1 335 315 discloses a technique for planning multiple non-diverse paths in a weighted non-directed graph.
• Dijkstra's algorithm is used twice; once for providing a routing tree from a start to all nodes of the graph and once for providing a routing tree specifically from a goal to all nodes. The weights from both trees are then summed to obtain the weight of going from the start to the goal via every node of the graph. Routes are tested for the ability to be topologically transformed into each other in the presence of obstacles so as to find topologically different paths. Such a technique is particularly suitable for robotic guidance.
• Yen's algorithm is not useful for finding good alternative routes, as the Kth shortest paths will simply be minor variations on the most optimum route. It works by repeatedly running Dijkstra's algorithm on a graph where edges and nodes are prohibited, by effectively setting their cost to infinity. In a road network, this is a non- starter for finding truly diverse routes as there is no way of knowing if the prohibited node or link is vital to one of the good alternative routes.
• Suurballe's algorithm and others are designed to find edge- or node-disjoint paths. These are unhelpful for finding good alternative routes because some road segments may be shared between several of the diverse alternative routes that we would like to find, which are therefore not disjoint. They work by running Dijkstra's algorithm to find a primary route, then altering all of the link costs, and running Dijkstra's algorithm again to find a backup route. For road route finding, we may not need such a mathematically rigorous definition of the K routes.
• US patent 6199009 discloses a technique for computing several routes based on the different preference settings (quickest, shortest, combination) and allowing the user to select between them.
• the problem with this approach is that in most cases the routes will be similar or even identical to each other (e.g. the shortest route may also be the quickest), and a radically different route that is perhaps 5% longer will just not be shown.
• Known systems guide the driver along one selected route without showing any en-route alternatives. Of course, if the driver strays from the selected route, then many commercial systems will recompute a new route, and US patent 5675492 shows how to precompute some alternatives and even display them. These will almost always be a short diversion back onto the earlier selected route, and are not interpreted as a choice for a completely different route.
• a method of generating a plurality of diverse routes from a source to a destination in a weighted directed graph comprising the steps of: generating a source routeing tree from the source to a first set of points of the graph; generating a destination routeing tree from a second set of points of the graph to the destination; and combining the source and destination trees to form the routes.
• the first set may comprise all of the points of the graph.
• the second set may comprise all of the points of the graph.
• the first and second sets may comprise points of the graph adjacent both the source and the destination.
• the first and second sets may comprise the same points.
• the graph may comprise a set of links, each of which has a first priority of use, and a second set of links, each of which has a second priority of use greater than the first priority.
• the first set of points may comprise points interconnected by the links of the first or second set in a first region of the graph containing the source and points interconnected by the links of the second set but not of the first set in a second region of the graph outside the first region.
• the second set of points may comprise points interconnected by the links of the first or second set in a third region of the graph containing the destination and points interconnected by the links of the second set but not of the first set in a fourth region of the graph outside the third region.
• the source and destination trees may be minimum cost trees.
• the combining step may comprise selecting each sub-route common to and traversed in the same direction by the source and destination trees.
• the source and destination trees may include back-pointers and the sub-routes may be selected by finding sequences of adjacent points which are pointed to by back-pointers of both the source and destination trees.
• the points of the source and destination trees may be associated with costs and the sub-routes may be selected by finding sequences of adjacent points for which the sums of the costs from the source and destination trees are the same.
• the combining step may further comprise extending each sub-route as necessary to the source and destination along the source and destination trees to form one of the routes.
• the method may further comprise assigning a measure of goodness to each route.
• the measure may be a function of the length of the sub-route and the length of the route.
• the measure may be a function of the difference between the length of the sub-route and the length of the route.
• the method may comprise selecting at least one via point which is not on any sub-route or extended sub-route, calculating from the source and destination trees a minimum cost route from the source via the at least one via point to the destination, and comparing the or each calculated cost with the same sum of the costs for at least one of the selected sub-routes.
• the method may comprise forming the difference between the or each calculated cost and the same sum for the selected sub-route.
• the method may comprise deselecting the or each via point for which the difference between the calculated cost and the same sum is greater than a threshold.
• the method may further comprise selecting only some of the routes in accordance with the measure of goodness.
• the method may comprise selecting the N routes of highest measure of goodness, where N is a positive integer.
• the measure may comprise selecting at least some of the routes whose measures of goodness are greater than a threshold.
• the method may comprise storing at least one of the routeing trees.
• the graph may comprise links associated with costs.
• the cost of traversing each of at least some of the links may vary with a parameter of traversal.
• the cost of traversing each of the least some of the links may vary with the time of traversal.
• the first of the generating steps to be performed may include calculating and storing the costs of the links for the prevailing transversal parameter values and the stored costs may be used during the second of the generating steps to be performed.
• the destination tree generating step may be performed before the source tree generating step.
• the routes may be ordered in accordance with at least one property of each route.
• the property may be the measure of goodness.
• the graph may represent a road network and the routes may be road routes.
• the graph may represent an integrated circuit or a printed circuit and the routes may be interconnections.
• the graph may represent a wiring installation and the routes may be wiring interconnections.
• the method may further comprise choosing the order of placement of the interconnections in accordance with the number of selected routes and/or the measure of goodness.
• the interconnections with the lowest number of selected routes may be placed first.
• the graph may represent a communications network and the routes may be communication paths.
• the network may be an internet.
• a method of navigation including generating road routes by a method according to the first aspect of the invention.
• the method may comprise presenting information about the road routes to a user.
• the method may comprise presenting information about a choice of routes from a point from which a plurality of the routes diverge when the user approaches the point.
• the information may be displayed in a form representing a road sign.
• the method may comprise offering guidance to a user along a selected one of the routes.
• the method may comprise offering guidance, when the user leaves the selected route, to the sub-route of the selected route.
• a computer program for performing a method according to the first or second aspect of the invention.
• a computer-readable medium carrying a program according to the third aspect of the invention.
• a computer programmed to perform a program according to the third aspect of the invention.
• a computer containing a program according to the third aspect of the invention.
• an apparatus arranged to perform a method according to the first or second aspect of the invention.
• Figure 1 illustrates a small hypothetical road network
• Figure 2 illustrates a source routeing tree generated by a method constituting an embodiment of the invention
• Figure 3 illustrates a destination routeing tree generated by the method
• Figure 4 illustrates the result of combining the routeing trees of Figures 2 and 3;
• Figure 5 illustrates the sub-routes or plateaux derived from Figure 4.
• Figure 6 illustrates an example of a route derived from Figure 4.
• Figure 7 illustrates a source routeing tree generated by the same or a similar method constituting an embodiment of the invention
• Figure 8 illustrates a destination routeing tree generated by the method
• Figure 9 illustrates the result of combining the routeing trees of Figures 7 and 8;
• Figure 10 illustrates the diverse routes generated by the method with ranking in accordance with a goodness factor;
• Figure 11 illustrates a source routeing tree found by the same or a similar method for a very simple graph
• Figure 12 illustrates a destination tree for the same graph
• Figure 13 illustrates the diverse routes generated by combining the routeing trees of Figures 11 and 12;
• Figure 14 illustrates an example of information display for displaying information about the generated routes to a user
• Figure 15 illustrates a navigation display which may be presented to a user
• Figure 16 illustrates how a road junction may be analysed within the method
• Figure 17 is a block schematic diagram of an apparatus constituting an embodiment of the invention.
• Figure 18 illustrates priority roads in a small hypothetical road network
• Figure 19 illustrates a source routing tree generated by a modified method constituting an embodiment of the invention
• Figure 20 illustrates a destination routing tree generated by the modified method
• Figure 21 illustrates the sub-routes or plateaux derived from combining the trees of Figures 19 and 20; and Figure 22 illustrates the diverse routes derived from the sub-routes of Figure 21.
• the first step is to compute the minimum cost tree from the source node to all other nodes. This is typically performed using a variant of Dijkstra's Algorithm or the A*
• each road segment we have annotated the distance from the source to that end using the shortest path. From each end of each road segment, there is exactly one outgoing arrowhead that shows the way back towards the source using the shortest path. We call this a back pointer. Note that for a source tree it will be in the opposite direction to the direction of travel. These are computed and stored as a necessary part of the Dijkstra's or A* Algorithm.
• the shortest path from the source to the node marked as the destination can be traced backwards by following the arrowheads back through the nodes whose distances are 282, 245, 207, 197, 130, 85, 25, 14 and finally, 0.
• This is the shortest path route from source to destination, and would be traced in just this way by Dijkstra's or the A* Algorithms.
• the second step is to compute the minimum cost tree to the destination node from all other nodes. This is just a variant of the previous algorithm and the output is shown as Figure 3. This time, the annotations give the distance to the destination node along the shortest path, and the arrowheads, again exactly one from the end of each road, show the way forward to the destination using the shortest path.
• the shortest path to the destination from the node at the top left with distance 234 is found by following the arrowheads through the adjacent nodes whose distances are 225, 215, 174, 113, 103, 65, 28, and 0.
• This tree also encodes the globally shortest path, which is found by following the arrowheads from the source node through to the destination node, and will always be identical to the one found from Figure 2.
• each node we begin by giving each node a single bit that indicates that we have visited it, initially 0 for not-visited. We then scan over every node in turn, the order does not matter. If it is marked as visited (1), we move on to the next node in the scan. If the node is marked as not- visited (0), then we change that to 1 to indicate that we have visited it. For each such newly-marked node, call it node Q, we begin a list of adjacent nodes in the chain by adding just a reference to that node Q.
• a plateau is formed when the source and destination trees traverse a chain of road segments in the same directions. This indicates that the chain is both useful for getting away from the source, and for getting towards the destination. This is a powerful indication that the chain may be useful for getting from the source to the destination.
• Such chains tend to use the best roads in their vicinity, and are aligned to help in getting from source to destination. On a real road network with millions of nodes, there are many thousands of such chains, many of them very short.
• a useful plateau for getting from source to destination will tend to be longer than those that are less useful, as it indicates a long stretch of route that is fast and well-aligned compared to others in its vicinity. Thus we are looking for a larger value of L.
• a useful plateau for getting from source to destination will tend to be part of a route that is not too long, as we are not interested in long plateaux if they are found at a great distance from both source and destination. Thus we are looking for a smaller value of SP + L + PD.
• the optimum route has a goodness G of 99
• routes where the route outside the plateau is about the same as the optimum route have a value of G around 1 (there are usually thousands of these, they are all the minor variations on the globally optimum route)
• worse routes have a value less than 1, which can go rapidly negative.
• G For presentation to the user, we can use the values of G as a threshold, and a value of G>50 gives good results. If they wish to see more routes, we can always keep more, down to some lower threshold, say G>10. Typically, this will be accomplished by the user asking for more alternatives, or altering the threshold value for G. This should not be necessary for the average user.
• G > 50 In testing on the complete Great Britain road network, from Cambridge to Canterbury, there is only one route with G > 50. This uses motorways for almost the entire length, namely the Mi l, M25 and M2. For purposes of investigation, we relaxed the threshold, but the routes with G ⁇ 50 were not at all appealing because those motorways are so well aligned and bounded by poor roads in London to the West and the sea to the East.
• the function that is used to find the minimum cost route can be much more complex that simply the distance travelled. In most situations, it is in fact dominated by the time taken, but may also be influenced, for example, by type of road, familiarity to the driver, financial costs, safety record in ways that are already known. We would typically use the best available time-independent cost function in the computation of the plateaux. It is simply a matter of labelling the roads with the appropriate cost instead of the distance, and the rest of the algorithm including the thresholding remains the same.
• Time dependent information could be used in the computation of the plateaux, but care would have to be taken that only one cost is associated with each road segment. Otherwise the source and destination trees might have different costs for the same segment, which would prevent us from identifying the chains using the Combined Cost (CC) values.
• CC Combined Cost
• An alternative method for handling time-dependent information is to compute the good alternative routes using the time-independent parts of the cost, and then compute the time-dependent factors for presentation to the user.
• Time dependent factors include the extra time taken due to congestion, variable toll costs and time-varying road charges.
• a user can see how the good geographical routes have been affected by the time-varying factors, and decide for themselves if a journey at another time might be more appropriate, or which route they would still like to select. For example, a driver's usual route may have become five minutes slower than the optimum due to congestion, but if they were given the choice, they may well wish to choose their usual route anyway. On another day, they might make the choice another way.
• the computer will never know enough to make this decision correctly under all circumstances, and so the driver should be informed about the realistic choices with the computer having done the hard work of estimating the distances, times, costs etc. of each choice.
• the table shows that with a threshold of G > 50, we would only recommend the routes generated from the plateaux with Combined Costs (CC) of 310, 332 and 335. Note that these happen to be the lowest values of the combined costs in the table, but that need not always be the case. Also note that all these routes have a cost that is within 10% of the globally optimum route, so other factors known only to the driver may easily outweigh the little extra distances involved, and they will be glad to have seen them all.
• CC Combined Costs
• the first step is to compute the tree containing the best (minimum-cost) routes from Cambridge to all other points. This is shown in Figure 7.
• the tree has been simplified for printing, showing only its major branches, as the leaves of the full tree would fill the land almost completely at this scale.
• Methods for computing such trees are well-known in the art, and include the Dijkstra and A* Algorithms.
• the second step is to compute the tree containing the best (minimum-cost) routes to Manchester from all other points. This is shown in Figure 8.
• the third step is to find roads that are traversed in the same direction in both trees, that is, away from Cambridge in Figure 7, and towards Manchester in Figure 8. Such roads are not simply the overlap of the two trees, as the direction of traversal is important.
• the resulting roads are shown in Figure 9.
• the fourth step begins by selecting the chains of roads from Figure 9 that are the longest, and generating complete routes from each by linking their endpoints back to Cambridge using the source tree of Figure 7, and back to Manchester using the destination tree of Figure 8.
• the routes are then ranked according to their goodness (the cost of the overlapping chain minus the cost of the entire route), and in this case, we have chosen to display the top five, numbered from 0 to 4 in order of goodness.
• the final step in this embodiment is to display relevant information to the user, such as that shown in Table 1.
• the routes generated from these chains are locally optimal, they are not globally optimal. There is, in general, only one globally optimal route, which is the one found by single routers such as Dijkstra's algorithm. The next most optimal route will usually be a minor variation upon the first, with only a tiny increase in cost. For example, on a three hour road journey that begins and ends in cities, there will be potentially hundreds of small deviations through side roads that add mere minutes to the computed journey time. All of their permutations will generate millions of routes with journey times up to, say three hours and ten minutes. Only after that might we see the first journey that used a very different route between the cities, and takes an extra 12 minutes.
• Dijkstra's algorithm In some implementations of known single routers (shortest single path from A to B), it may be desirable to use a variation of Dijkstra's algorithm. These use special heuristics to explore the nodes or edges in a better order, or to terminate earlier, or to explore major roads only, or to explore outwards from source and destination simultaneously. They do this in order to run faster or to use less memory. It should be understood that our method of combining RTs to find plateaux is independent of how the RTs were computed, so if these variations are appropriate, then our method can be applied to them to yield diverse alternative routes while taking advantage of their better running time or memory use characteristics. Cost Function
• the cost function that is used to compute the best diverse alternative routes can incorporate any of the known factors. It may be sensitive to both time and duration, can incorporate real-time or historical traffic information, can take into account user preferences, and can use financial information such as road pricing.
• Another problem with known route planners is that the user may wonder why some other route was not recommended. By showing them the alternative routes along with their characteristics, it is easy for the user to see how their alternative route compares to the best, and to make an appropriate choice. For example, the usual route may be subject to unusually high congestion, or an alternative route may have become preferable because of the construction of a bypass.
• a route planning tool we could highlight the different routes using different colours, and present a text box with relevant details about each route such as time, cost etc.
• An example of such a text box for the Cambridge to Manchester routes described hereinbefore is shown in Figure 14. Selecting a route on the map could highlight the relevant line in the text box, and selecting a line in the text box could highlight the relevant route on the map. In this way, a user could make their choice of route with as much information as we can give.
• Destination Link We begin with the Source Link, which is found at the beginning of every Choice Route. If the next link on every Choice Route is the same, then we move on to that. If the next link is different on different routes, then we have reached a choice point, and follow each route separately as if we had just begun again at the source link. We finish when each route arrives at the Destination Link. We can either compute all of the Choice points as we begin the journey, or we can compute them as we go, always keeping at least one known in advance.
• junctions For road networks for example, we may represent junctions as nodes in the graph, and road links between them as edges in the graph.
• costs of traversing the roads for example the distance and average time taken to traverse a length of road between junctions
• costs to the edges For represent the costs of passing through a junction, for example the time taken to proceed straight on, or to turn left or right, we can assign costs to a node that are dependent on the edge used to reach the node, and the edge used to leave that node.
• each edge is uni-directional, so a stretch of two-way road is represented by two edges, one for each direction of travel, and these strictly form the nodes of the graph. This allows us to encode different delays for each direction of travel, which may be important at times of congestion.
• each edge a Link.
• each Link has a set of pointers to the next Links that can be reached from it in the graph. These are called the NextLinks.
• NextLink pointers are strictly the edges of the graph. Associated with each such pointer are the costs (in time, distance, financial cost etc.) of moving from the original Link to the next one. Our implementation chooses to use the time taken to move from the midpoint of one Link edge to the midpoint of another (NextLinkTime), but others may choose to use the beginning of the Links, or their ends, it makes no difference to the outcome.
• NextLinkTime the time taken to move from the midpoint of one Link edge to the midpoint of another
• NextLinkTime the time taken to move from the midpoint of one Link edge to the midpoint of another
• each Link For computing optimum routes to a particular destination Link rather than from a source Link, we give each Link a set of PrevLinks, which encode the costs of entering that Link from all possible predecessor Links, that is, the segments of road from which it can be entered. There is some redundancy here, as every Prevlink has a corresponding NextLink, so we use more memory than some representations, but we have a lower running time.
• Figure 16 shows a picture of a road junction, where the side road is one-way towards the junction.
• the section of road marked A is represented by two links, marked 1 and 3, as it is a two-way road.
• road segment B is represented by two links, 2 and 4.
• Road segment C is only represented by one link, 5, as it is a one-way road towards the junction.
• NextLink (arrow) emanating from link 1, which points to link 2.
• NextLink emanating from link 4 which points to link 3.
• Vehicles travelling on road segment C towards the junction can turn onto either segment A or segment B travelling away from the junction, so the corresponding link 5 has two NextLinks emanating from it, pointing to links 2 and 3.
• An important factor here is that the NextLinks can be enumerated from the link from which they emanate, not from the links that they point towards.
• the PrevLinks are the inverse of the NextLinks, and are again enumerated from the link that they emanate from. They indicate the links from which vehicles may enter this link. For example, vehicles may enter road segment A at this junction from either road segment B or C, so link 3 has PrevLinks that point to link 4 and to link 5.
• This link data is stored as a set of files, one for each type of information, that can easily be mapped into memory for efficient access.
• Each link is given a simple integer ID, length 32 bits, from 0 upwards, which is then used as an offset into various files.
• linkID 435 as an example.
• the length of link 435 will be found at the 435 th 16-bit word of file LinkDistance, encoded as the length in metres in 16 bits.
• the links that can be reached from link 435, its NextLinks have 32-bit linkIDs in a file called NextLinks.
• the 435 th 32-bit word in file NextLinklndex gives the offset in the file NextLinks where the linkIDs of the next links will be found, contiguously.
• nLinks nNextLinks.
• INT32 Contains the number of links (nLinks) and the number of next-links (nNextLinks), and prev-links (nPrevLinks) and maybe other top level information about how the files were generated etc. So far, all three fields are INT32.
• a link is defined by its link index, an integer between 0 and nLinks- 1, which indexes all of the arrays contained in files whose length is a multiple of [nLinks].
• PrevLink encodes the routeing tree for destination routeing.
• the routeing algorithm that we use (a version of Dijkstra's algorithm) to construct the min-cost routes from a given link to all other links (the Source Routeing Tree) is as follows:
• TTBacklndexl contains the back pointers for tree 1.
• n is 0 and 1.
• file lengths are given in square brackets after the file name:
• TTBackIndex ⁇ n> [nLinks]: INT32: The link that was used to get to this link at lowest cost, for tree ⁇ n>. Works for Next or Prev (source or destination) trees.
• the link that led us to linkB along the optimum route is found at offset 678,123 in the file TTBacklndexO.
• this link had ID 1,456,789.
• the link that led us to this one along the optimum route is found at offset 1,456,789 in the file TTBacklndexO.
• the routeing algorithm that we use (a version of Dijkstra's algorithm) to construct the min-cost routes from all links to a given link (the Destination Routeing Tree) is as follows:
• TTBacklndexl file For example, suppose that linkA had ID 238,345. The link that linkA preceded along the optimum route is found at offset 238,345 in the file TTBacklndexl. Suppose this link had ID 1,432,876. The link this one preceded along the optimum route is found at offset 1,432,876 in the file TTBacklndexl. We keep repeating this until the link ID extracted from the file matches the linkID of the destination link linkB. Then we have extracted all of the link IDs of the optimum (shortest) path in forward order. This is a common way of extracting the shortest path once Dijkstra's algorithm has terminated.
• this path will be identical to the one extracted from the Source Routeing Tree in the section above. It does not matter if Dijkstra's algorithm is run from the source node outwards or the destination node inwards, it will always find the single shortest path route from linkA to linkB. However, the Routeing trees that are a by-product of Dijkstra's algorithm are very different. The only route that they share completely in common is the global shortest-path route.
• the cost at offset ⁇ linkC> in the file TTMinCostl is the cost of the minimum cost route from linkC to the destination linkB.
• the cost at offset ⁇ linkC> in the file TTMinCostO is the cost of the minimum cost route from the source link linkA to linkC. If we add these two quantities together, we have found the cost of the minimum cost route from linkA via linkC to linkB.
• the chains are identified by the cost of going from linkA to linkB on mincost routes via any of the links in the chain being the same cost. For this reason, we call such a chain a plateau, as the sum of the costs is constant along the links that it is composed of.
• next two subsections give alternative methods for finding the links that are part of plateaux, which may have different advantages according to the properties of the cost function, and the memory and processor architecture on which the methods are implemented. It will also be apparent that these methods can be adapted to different graph representations, means of storing the flags, and properties of the cost functions.
• TCFlags01[nLinks] A series of INT8's used as flags when combining RTs denoted by O and 1.
• n goodness values where n is greater than the number that we could possibly want to display to the user, say 100.
• the goodness could be computed in any way, using any of the properties of routes such as length, financial cost, types of road, distance from source or destination.
• raw goodness is the negative of the cost of the route involving the plateau that is not part of the plateau, divided by the cost of the globally mincost route (GMCost).
• This raw goodness will have a value of zero for the globally minimum cost route, as there is zero cost outside the plateau.
• goodness 100 - (99 aw - 8 °° )
• the set of diverse alternative routes should include all those that we knew about as good routes from our own experiences, or at least minor variations on them (as the routeing database does not perfectly reflect the user's view of costs, but is roughly correct over the longer stretches).
• the set of diverse alternative routes should include each of the best single routes computed by other route planners such as (for Great Britain) Microsoft AutoRoute, ALK CoPilot, Medion Navigator, ISYS Personal Navigator.
• the trees can be saved by keeping the files in which their costs and back pointers have been computed, or if an alternative implementation has been used, then new file formats could be defined for the storage of this information. If storage is at a premium, then strictly only one of either the back pointers or the costs need to be stored, as a scan through the routeing database can be used to reconstruct the other without the need to run a full routeing algorithm again.
• a conventional router will run something similar to Dijkstra's algorithm to find the best route, where the link costs have been altered to reflect the latest time-dependent information, and where they are sensitive to the time that the links are expected to be traversed.
• the choice routes computed under the restricted cost function may no longer include the globally optimum route under the full cost criteria.
• time-dependent optimal routes are likely to involve more minor roads, along which there are no traffic sensors or reports. These roads appear to be free-flowing, where in reality they may be clogged, so the chosen routes may not be optimum after all.
• the RTs using the basic cost function do not have to be re-computed when the other cost factors are varied.
• a good desktop interface should allow the user to sort the alternative routes by time, distance, financial cost, type of road, number of turns etc. as they explore their options. This does not necessarily require any recomputations, as those factors can be computed and stored along with the alternative routes when they are first generated.
• the Choice Routes do not have to be used singly. If we have journey information in the form of Origins and Destinations of journeys, we can use the choice routes to allocate traffic to all of the links used in those routes in proportion to the goodness of the routes.
• Methods of this type may also be applied to the routeing of wires, conductors or cables, for example in buildings, aircraft, printed circuit boards and integrated circuits.
• finding a route for a wire there are a few differences from the case for road routes, particularly the routeing of wiring nets, and the additional variable of placement order.
• this technique does not replace the known techniques, as we may still have to rip-up and re-route some nets, but we will sort the unplaced nets by the amount of choice available, always placing those with the least choice first, which should speed up the time to find a solution, and more importantly, will improve the final result by keeping the wiring nets shorter overall, or not requiring such wide routeing channels.
• Figure 17 shows an example of a navigation apparatus in the form of a computer 10 with a program memory 11 , such as a ROM, flash memory, a hard disk drive and/or an optical disk drive, installed in a vehicle such as an automobile.
• a program memory 11 such as a ROM, flash memory, a hard disk drive and/or an optical disk drive
• a vehicle such as an automobile.
• a program memory 11 such as a ROM, flash memory, a hard disk drive and/or an optical disk drive
• a vehicle such as an automobile.
• a program memory 11 such as a ROM, flash memory, a hard disk drive and/or an optical disk drive
• a vehicle such as an automobile.
• such an apparatus is not limited to use in vehicles and may be attached to or installed in other objects.
• such an apparatus may be installed in a cellular or mobile telephone, for example of the type which receives information about its geographical location and about the current time.
• the vehicle comprises vehicle electronics 12, which monitor and control many of the systems on board the vehicle.
• the computer 10 is thus able to receive information about various aspects of the current state of the vehicle.
• the vehicle is further provided with a GPS receiver 13 for determining the location of the vehicle and the current time from signals received from the GPS via an aerial 14. This information is also supplied to the vehicle electronics 12, for example so as to provide the function of a satellite navigation or "Sat Nav" system.
• the computer 10 is also provided with an output device 15 for providing a human- perceivable output.
• the output device 15 is illustrated in the form of a display, for example forming part of the Sat Nav system or of an in- vehicle entertainment system, but may alternatively or additionally comprise an audio output device such as a loudspeaker.
• a common means of making route computations faster is to restrict the search to a subset of roads that are generally faster or more useful when sufficiently far from the origin or destination of the journey.
• priority roads and the subset is arranged to also include junctions, slip roads etc. to ensure that they are fully connected.
• This method has two advantages over trying to compute a source or destination tree with slower roads used at both source and destination. Firstly, we reduce the computation overheads from two slow road type searches in each tree to just one for each tree, so choice routing becomes more efficient and closer to the costs of just running a conventional router. Secondly, the decision on when to switch from slow roads to a faster road type is much easier when moving away from the origin in the source tree, or away from the destination in the destination tree. This is because we can notice how many roads of the higher types we have encountered and use that as part of the switching criteria. This can easily be done several times for a hierarchy of faster road types.
• a conventional router would have higher overheads when moving towards the destination in the source tree, because of the occasional difficult case where slower road types must be used for the last 30 miles, but which is not known in advance. Thus, the conventional router would switch at this worst case distance all of the time. For the common cases though, where slower road types might only be necessary for the last 5 miles, switching to slower road types at 30 miles distance is very wasteful.
• time dependent information such as the average road speeds for each five minute period of each day
• junctions at which one or more choices diverge we can also compute the junctions at which one or more choices diverge. As described hereinbefore, we can present these choices along with information about their properties to the driver well in advance of the junction so that the driver can simply choose which way they want to go and drive along it. If the driver has requested route guidance, how should that work at the choice point? Presentation of the choices as a road sign is a good substitute for the usual junction diagram with one exit highlighted, and the audio guidance could similarly offer two or more options as this is being done well in advance of the junction.
• the driver could now choose a route that they wished to be guided on by selecting one of the routes, by some conventional interface such as a button, touchscreen, or by voice command.
• some conventional interface such as a button, touchscreen, or by voice command.
• a conventional guidance system would recompute the optimal route from the current position to the destination and resume guidance along that. Our problem is that this new optimal route may not be the chosen choice route, so the driver would find themselves not being guided along the way that they had chosen. Once the driver was far enough along the chosen route that they were on the plateau, then we know that the optimal route to the destination is indeed the choice route itself but, until the route reaches the plateau section, there will in general be a more optimal route that does not use the plateau. This then gives us one possible solution.
• any of the conventional solutions can be used. For example, if the driver strays from the route that we are guiding them along and the newly computed optimal route to the via becomes one that moves away from the final destination for some distance (or cost), as known from the destination tree, then we can cancel the guidance to the via as if it had been reached and continue guidance from the next choice point after that via or the beginning of the plateau on the route with the highest goodness, whichever would have been reached first.
• An example of this is where we want to look for good stopping places on a journey.
• we can find all points of interest perhaps restaurants, toilet facilities, petrol stations, rest areas) that lie within some distance from it as the crow flies.
• One technique is to compute the perpendicular distance to each link on the route to find the minimum perpendicular distance, and another is to compute the straight-line distance to each junction on the route to find a lower bound on the driving distance.
• the trees can also contain the cumulative time, distance, safety or other factors, in addition to the cost function that they are minimising, and then these can be used in the thresholding. For example, we may find all those restaurants that do not add more than ten minutes to our estimated driving time. To do this without using the combined tree would involve running a routing algorithm twice for each node found by the proximity search, which would complicate the code, take an uncomfortably long time, and consume precious CPU and memory resources.

Landscapes

• Engineering & Computer Science (AREA)
• Business, Economics & Management (AREA)
• Physics & Mathematics (AREA)
• Radar, Positioning & Navigation (AREA)
• Remote Sensing (AREA)
• General Physics & Mathematics (AREA)
• Theoretical Computer Science (AREA)
• Strategic Management (AREA)
• Development Economics (AREA)
• Economics (AREA)
• Human Resources & Organizations (AREA)
• Game Theory and Decision Science (AREA)
• Computer Networks & Wireless Communication (AREA)
• Entrepreneurship & Innovation (AREA)
• Marketing (AREA)
• Finance (AREA)
• Accounting & Taxation (AREA)
• Signal Processing (AREA)
• General Business, Economics & Management (AREA)
• Automation & Control Theory (AREA)
• Tourism & Hospitality (AREA)
• Mathematical Physics (AREA)
• Educational Administration (AREA)
• Educational Technology (AREA)
• Quality & Reliability (AREA)
• Operations Research (AREA)
• Navigation (AREA)

Priority Applications (3)

Applications Claiming Priority (2)

Publications (1)

Family

ID=37546166

Family Applications (1)

Country Status (5)

Cited By (7)

Families Citing this family (130)

Citations (4)

Family Cites Families (12)

• 2006
  • 2006-10-30 GB GB0621508A patent/GB2443472A/en not_active Withdrawn
• 2007
  • 2007-09-18 JP JP2009535133A patent/JP5448827B2/ja not_active Expired - Fee Related
  • 2007-09-18 WO PCT/GB2007/050558 patent/WO2008053240A1/en not_active Ceased
  • 2007-09-18 US US12/445,238 patent/US8249810B2/en not_active Expired - Fee Related
  • 2007-09-18 EP EP07804462.5A patent/EP2084493B1/en active Active

Patent Citations (4)

Also Published As

Similar Documents

Legal Events