CN114815834A - Dynamic path planning method for mobile intelligent agent in stage environment - Google Patents
Dynamic path planning method for mobile intelligent agent in stage environment Download PDFInfo
- Publication number
- CN114815834A CN114815834A CN202210465123.7A CN202210465123A CN114815834A CN 114815834 A CN114815834 A CN 114815834A CN 202210465123 A CN202210465123 A CN 202210465123A CN 114815834 A CN114815834 A CN 114815834A
- Authority
- CN
- China
- Prior art keywords
- dynamic
- state
- obstacle
- agent
- mobile intelligent
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 38
- 230000003068 static effect Effects 0.000 claims abstract description 26
- 238000004088 simulation Methods 0.000 claims abstract description 16
- 230000007246 mechanism Effects 0.000 claims abstract description 10
- 230000006870 function Effects 0.000 claims description 26
- 230000033001 locomotion Effects 0.000 claims description 15
- 238000013528 artificial neural network Methods 0.000 claims description 14
- 230000004888 barrier function Effects 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 10
- 230000009471 action Effects 0.000 claims description 9
- 230000007704 transition Effects 0.000 claims description 8
- 230000003993 interaction Effects 0.000 claims description 7
- 230000015654 memory Effects 0.000 claims description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 230000001186 cumulative effect Effects 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 2
- 230000002452 interceptive effect Effects 0.000 claims description 2
- 230000000644 propagated effect Effects 0.000 claims description 2
- 241000283283 Orcinus orca Species 0.000 claims 1
- 230000000694 effects Effects 0.000 description 5
- 230000002787 reinforcement Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0223—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving speed control of the vehicle
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0221—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving a learning process
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0276—Control of position or course in two dimensions specially adapted to land vehicles using signals provided by a source external to the vehicle
Landscapes
- Engineering & Computer Science (AREA)
- Aviation & Aerospace Engineering (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a dynamic path planning method for a mobile intelligent agent in a stage environment, belonging to the technical field of path planning of intelligent robots; the method comprises the steps of firstly, obtaining obstacle information around a mobile intelligent body by constructing a global map, classifying obstacles into dynamic obstacles and static obstacles, then establishing a local map, coding the dynamic obstacle information through an LSTM network, and calculating the importance of each dynamic obstacle through a social attention mechanism to achieve better obstacle avoidance. Different avoiding conditions of the dynamic and static obstacles are responded by constructing a new reward function, so that the path planning problem of the mobile intelligent body under the complex stage environment is realized. A new experience pool updating method is provided to improve the convergence speed of network training, and meanwhile, simulation experiments are carried out on the method provided by the invention to prove the superiority of the algorithm, so that the method has very high practical value.
Description
Technical Field
The invention relates to the technical field of intelligent robot path planning, in particular to a dynamic path planning method for a mobile intelligent agent in a stage environment.
Background
In order to meet the demand of basic-level culture diversified services, culture service functions such as culture performance, meeting agenda, exhibition reading, folk custom activities and the like need to be developed in the activity space of the basic-level small-sized culture service complex. The cultural facilities with classified configuration cannot meet the expectation of residents on comprehensive cultural demands. The small culture service complex can better solve the problem, has the key point of promoting the construction of multifunctional activity space, can integrate the functions of folk-custom activities, exhibition, meeting, reading and the like to form an integrated venue service carrier, not only meets the requirements of diversity and self-organization of rural basic culture, but also explores a new mode meeting the requirements of new rural public culture service in China.
In order to reduce waste of land resources and improve space utilization efficiency, various intelligent moving bodies are needed to assist in completing mutual rapid combination and switching of multiple functional spaces, configuration facilities and use requirements in the functional spaces are different due to different service spaces, in order to meet the requirement of one hall with multiple purposes of a small-sized cultural complex, the intelligent moving bodies are often used for assisting in completing mutual rapid combination and switching of the multiple functional spaces, dynamic path planning and the like in a crowded environment in the space of the small-sized complex are realized, and accordingly multiple cultural service requirements of a single space are met.
The small-sized cultural complex space is a typical human, machine and object coexistence environment, in the functional space switching process, trajectory planning movement is carried out on a plurality of object devices in the space under the condition that other devices are needed, and the switching service of the space function is realized, so that how to quickly avoid dynamic and static obstacles in the switching process and reach a target point needs to design a dynamic path planning algorithm to control an intelligent moving body, and the cultural complex space is crowded in the environment and has higher requirements on the dynamic path planning algorithm.
The traditional dynamic path planning algorithm relies on the quick refreshing of a sensor to sense the information of surrounding obstacles, and the planned path also has the problem of detour or unnatural track along with the change of the dynamic obstacles. The movement tendency of surrounding dynamic obstacles cannot be predicted, and adaptability is poor. In view of the above problems, it is desirable to provide a dynamic path planning method capable of distinguishing dynamic and static obstacles and predicting dynamic obstacle trends.
Disclosure of Invention
Aiming at the problems in the prior art, the invention aims to provide a dynamic path planning method for a mobile intelligent body in a stage environment, which designs a new Markov decision process and a network structure on the basis of a deep reinforcement learning method, adds an attention score to a dynamic obstacle by introducing a social attention mechanism, solves the problem of unfixed dimension of a feedforward network by using a long-short term memory neural network, constructs a new reward function to deal with different avoidance conditions of the dynamic and static obstacle, and provides a new experience pool updating method to improve the convergence speed of network training, so that the mobile intelligent body realizes the distinguishing of the dynamic and static obstacle and the prediction of the motion trend of the dynamic obstacle. In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
a dynamic path planning method for a mobile intelligent agent in a stage environment comprises the following steps:
1) establishing a simulation environment model of the mobile intelligent body and the dynamic and static barriers based on the gym library;
2) designing a Markov decision process, and designing a state space S, an action space A, a transition probability P, an award R and a discount factor gamma;
3) designing a neural network structure;
4) initializing network parameters by emulating learning pretrains using an optimal reciprocal collision avoidance algorithm (ORCA); after the simulation learning is finished, training is carried out through the actual interaction of the mobile intelligent agent in the simulation environment to optimize network parameters;
5) training a neural network by an adaptive moment estimation method (Adam) to obtain an optimal value function:
V * (u t )＝∑γ Δt·Vpref ·P(u t ,a t )
6) setting an optimal strategy by maximizing cumulative returns:
wherein u is t Indicating the joint status of the current mobile agent and the obstacle, a t Representing a set of motion spaces, gamma a decay factor, Δ t a time interval between two motions, Vpref a preferred speed, V * Expressed in the optimal value function, P is expressed as a state transition function, and R is expressed as a reward function;
indicating the joint state at the next time instant
7) Selecting an action a at the current moment according to an optimal strategy t Until the mobile agent reaches the target.
Further, the mobile agent and the dynamic obstacle in step 1) are set to be circles with a radius of 0.3 m, and the static obstacle is defined as a circle with a radius of 0.5 m to 1 m or a quadrangle with an area of 1 m to 1.5 m.
Further, in the step 2), a state space S is set, wherein the state of the dynamic obstacle is S D ＝[P x ,P y ,V x ,V y ,r,V pref ]The state of the static obstacle is S S ＝[P x ,P y ,r]The state of the mobile agent is S T ＝[P x ,P y ,G x ,G y ,V x ,V y ,θ,r,V pref ]Joint state u t ＝[S T ,S S ,S D ]Wherein (P) x ,P y ) To move the current position of the agent and the dynamic and static obstacles, (G) x ,G y ) Is the position of the set target point, theta is the course angle of the mobile intelligent body, r is the radius of the mobile intelligent body and the dynamic and static barriers, and V pref (V) preferred speed for moving agent x ,V y ) The moving speed of the moving intelligent agent and the dynamic barrier;
the motion space A is linear velocity and angular velocity, and in order to meet dynamic constraint, the angular velocity is divided into 18 equal parts of [ - π/4, π/4]Within interval, linear velocity is as a function
x is 1, 2, 3, 4, 5 to obtain 5 changesThe smooth linear velocity is changed, and the motion space has 90 motion combinations;
the transition probability P is approximately calculated through a track prediction model;
the reward R is set as:
wherein G x,y Is the position information of the target point, P x,y Is the current location information of the mobile agent, d s Is the distance between the mobile agent and the static obstacle, d d Is the distance between the mobile agent and the dynamic barrier; the discount factor gamma is taken to be 0.9.
Further, the network structure in step 3) is composed of the following modules: 1. an input layer: the input layer is the joint state u in the above steps t ＝[S T ,S S ,S D ]. 2. Long short term memory neural network module (LSTM): obstacles around the mobile agent may be sorted by the LSTM module and network layer output parameters may be fixed. 3. Social attention mechanism: the probability of collision between the mobile agent and surrounding dynamic obstacles can be analyzed through a social attention mechanism module, and the probability is displayed in a score form. 4. An output layer: the output layer outputs an optimal value function V through weighted linear combination of network parameters * (u t )。
Further, the network operation flow in step 3) is as follows: firstly, state information of a mobile intelligent body and an obstacle is input into a network, then the obstacle is divided into a dynamic obstacle and a static obstacle according to the state information, the state of the mobile intelligent body and the state of the dynamic obstacle are input into an LSTM module, then the state of the mobile intelligent body and the state of the dynamic obstacle are input into a social attention mechanism module, the processed state, the obtained interactive characteristic and the state of the static obstacle are input into two layers of full-connected layers, and finally the state is normalized through an activation function to obtain an optimal value function.
Further, in the step 4), when the mobile agent interacts in the simulation environment, the current state information, the action information and the reward information are stored in the experience pool as one experience, and when the experience reaches the maximum capacity, the new experience is stored instead of the old experience with low reward, so that the probability of selecting excellent experience is improved, and the convergence rate of the network is improved. The current set is ended when the mobile agent encounters an obstacle or exceeds the maximum time for a single set run during each set interaction. The experience is then propagated through the gradient directions to update the network parameters.
The invention has the beneficial effects that: a new Markov decision process is designed to adapt to the complex situation of the obstacles in the complex space, a new network structure is designed to realize classification processing of dynamic and static obstacles and prediction of dynamic obstacles, a new reward function is designed to cope with the situations of different obstacles, a new experience pool updating method is provided to improve the training efficiency of a neural network, the network is pre-trained by using simulation learning, and the convergence speed of the network is improved; therefore, the mobile intelligent agent can realize an efficient dynamic planning method in the complex space.
Drawings
FIG. 1 is a flow chart of a method implementation of an embodiment of the present invention;
FIG. 2 is a diagram of a network architecture in an embodiment of the present invention;
FIG. 3 is a graph of simulation results in an embodiment of the present invention;
FIG. 4 is a network training total rewards graph in an embodiment of the invention.
Detailed Description
The present invention will be described in further detail with reference to the following drawings and specific embodiments.
The invention aims to provide a dynamic path planning method of a mobile intelligent body in a stage environment, which designs a new Markov decision process and a network structure on the basis of a deep reinforcement learning method, adds an attention score to a dynamic barrier by introducing a social attention mechanism, solves the problem of unfixed dimension of a feedforward network by using a long-short term memory neural network, constructs a new reward function to deal with different avoidance conditions of the dynamic barrier and the static barrier, and provides a new experience pool updating method to improve the convergence speed of network training, so that the mobile intelligent body can distinguish the dynamic barrier and predict the motion trend of the dynamic barrier.
In this embodiment, as shown in fig. 3, the simulation environment is set to have a planning map range of 10 × 10, a starting point of the path planning is (0, -10), a target point is (0,7.5), positions of the static obstacles are randomly distributed and are rectangles or squares, and the dynamic obstacles are circles with a radius of 0.5.
A dynamic path planning algorithm for a mobile intelligent agent in a stage environment comprises the following specific steps:
1) establishing a simulation environment model of the mobile intelligent body and the dynamic and static barriers based on the gym library;
2) designing a Markov decision process, and designing a state space S, an action space A, a transition probability P, an award R and a discount factor gamma;
3) designing a neural network structure;
4) initializing network parameters by emulating a learning pretraining 3000 set using an optimal reciprocal collision avoidance algorithm (ORCA); and then training to optimize network parameters through the actual interaction of the mobile agent in the simulation environment.
5) Training a neural network by an adaptive moment estimation method (Adam) to obtain an optimal value function:
V * (u t )＝∑γ Δt·Vpref ·P(u t ,a t )
6) setting an optimal strategy by maximizing cumulative returns:
wherein u is t Representing the joint state of the current moving agent and the obstacle, a representing a set of motion spaces, γ representing a decay factor, Δ t representing a time interval between two motions, Vpref representing a preferred velocity, V * Expressed in the optimal value function, P is expressed as a state transition function, and R is expressed as a reward function;
indicating the joint state at the next time instant
7) Selecting an action a at the current moment according to an optimal strategy t Until the mobile agent reaches the target.
The above are preferred embodiments of the present invention, and all changes made according to the technical scheme of the present invention that produce functional effects do not exceed the scope of the technical scheme of the present invention belong to the protection scope of the present invention.
Claims (6)
1. A dynamic path planning method for a mobile intelligent agent in a stage environment is characterized by comprising the following steps:
1) establishing a simulation environment model of the mobile intelligent body and the dynamic and static barriers based on the gym library;
2) designing a Markov decision process, wherein the Markov decision process represents < S, A, P, R, gamma > through a quintuple, and designs a state space S, an action space A, a transition probability P, a reward R and a discount factor gamma;
3) designing a neural network structure;
4) initializing network parameters by simulating learning pretrains using an optimal reciprocal collision avoidance algorithm ORCA; after the simulation learning is finished, training is carried out through the actual interaction of the mobile intelligent agent in the simulation environment to optimize network parameters;
5) training a neural network by an adaptive moment estimation method Adam to obtain an optimal value function:
V * (u t )＝∑γ Δt·Vpref ·P(u t ,a t )
6) setting an optimal strategy by maximizing cumulative returns:
wherein u is t Indicating the joint status of the current mobile agent and the obstacle, a t Representing a set of motion spaces, gamma representing a decay factor, and at representing a time interval between two motionsVpref represents the preferred speed, V * Expressed in the optimal value function, P is expressed as a state transition function, and R is expressed as a reward function;
representing the joint state at the next time;
7) selecting an action a at the current moment according to an optimal strategy t Until the mobile agent reaches the target.
2. A dynamic path planning method for mobile agents in stage environment as claimed in claim 1, wherein in step 1), a simulation environment model is built based on the gym library, the mobile agents and the dynamic obstacles are set to be circles with a radius of 0.3 m, and the static obstacles are defined to be circles with a radius of 0.5 m to 1 m or quadrangles with an area of 1 m to 1.5 m.
3. A stage environment mobile intelligent dynamic path planning method according to claim 1, wherein in step 2), a state space S is set, wherein the state of the dynamic obstacle is S D ＝[P x ,P y ,V x ,V y ,r,V pref ]The state of the static obstacle is S S ＝[P x ,P y ,r]The state of the mobile agent is S T ＝[P x ,P y ,G x ,G y ,V x ,V y ,θ,r,V pref ]Union state u t ＝[S T ,S S ,S D ](ii) a Wherein (P) x ,P y ) To move the current position of the agent and the dynamic and static obstacles, (G) x ,G y ) Is the position of the set target point, theta is the course angle of the mobile intelligent body, r is the radius of the mobile intelligent body and the dynamic and static barriers, and V pref (V) preferred speed for moving agent x ,V y ) The moving speed of the moving intelligent agent and the dynamic barrier;
the motion space A is linear velocity and angular velocity, in order to accord with dynamicsConstrained, angular velocity split of 18 equal divisions to [ - π/4, π/4]Within the interval, the linear velocity is according to an exponential function
x is 1, 2, 3, 4 and 5, so that 5 smoothly-changing linear velocities can be obtained; the motion space has 90 motion combinations;
the transition probability P is used for transferring the state through the actual interaction of the mobile intelligent agent in the simulation environment; the reward R is set as:
wherein G is x,y Is the position information of the target point, P x,y Is the current location information of the mobile agent, d s Is the distance between the mobile agent and the static obstacle, d d Is the distance between the moving agent and the dynamic barrier.
4. A stage environment mobile intelligent dynamic path planning method according to claim 1, wherein the neural network structure in step 3) includes: the system comprises an input layer, a long-short term memory neural network module (LSTM), a social attention mechanism and an output layer;
wherein the input layer: the input layer is the joint state u in the above steps t ＝[S T ,S S ,S D ](ii) a Long-short term memory neural network module LSTM: sorting obstacles around the mobile agent through an LSTM module, and fixing network layer output parameters; social attention mechanism: analyzing the probability of collision between the mobile intelligent agent and surrounding dynamic obstacles through a social attention mechanism module, and displaying the probability in a score form; an output layer: the output layer outputs an optimal value function V through weighted linear combination of network parameters * (u t )。
5. A stage environment mobile intelligent dynamic path planning method according to claim 1, wherein the network operation flow in step 3) is as follows: firstly, state information of a mobile intelligent body and an obstacle is input into a neural network structure, then the obstacle is divided into a dynamic obstacle and a static obstacle according to the state information, the state of the mobile intelligent body and the state of the dynamic obstacle are input into an LSTM module, then the state of the mobile intelligent body and the state of the dynamic obstacle are input into a social attention mechanism module, the processed state, the obtained interactive characteristic and the state of the static obstacle are input into two layers of full-connected layers, and finally the state is normalized through an activation function to obtain an optimal value function.
6. A dynamic path planning method for a mobile agent in a stage environment according to claim 1, wherein in step 4), the mobile agent stores current state information, action information and reward information as an experience in an experience pool during interaction in a simulation environment, and importance is added to each experience through TD-error, TD-error is a difference between a value function of an action at a certain moment and an optimal value function of a current network, and if the difference is larger, the current experience is worse; by definition:
P t ＝(|δ t |+ε) α
wherein P is t Is to select the probability of the current experience, alpha and epsilon are constants, delta t For TD-error, ε is to prevent the empirical TD-error from being replayed after 0;
different probabilities are given to experiences according to different experiences, so that the probability of selecting excellent experiences is improved, and the convergence speed of the network is improved; ending the current set when the mobile agent encounters an obstacle or exceeds the maximum time of single set operation in the interaction process of each set; the experience is then propagated back through the gradient to update the network parameters.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210465123.7A CN114815834A (en) | 2022-04-29 | 2022-04-29 | Dynamic path planning method for mobile intelligent agent in stage environment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210465123.7A CN114815834A (en) | 2022-04-29 | 2022-04-29 | Dynamic path planning method for mobile intelligent agent in stage environment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114815834A true CN114815834A (en) | 2022-07-29 |
Family
ID=82509534
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210465123.7A Pending CN114815834A (en) | 2022-04-29 | 2022-04-29 | Dynamic path planning method for mobile intelligent agent in stage environment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114815834A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116090688A (en) * | 2023-04-10 | 2023-05-09 | 中国人民解放军国防科技大学 | Moving target traversal access sequence planning method based on improved pointer network |
| CN118394109A (en) * | 2024-06-26 | 2024-07-26 | 烟台中飞海装科技有限公司 | Simulated countermeasure training method based on multi-agent reinforcement learning |
-
2022
- 2022-04-29 CN CN202210465123.7A patent/CN114815834A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116090688A (en) * | 2023-04-10 | 2023-05-09 | 中国人民解放军国防科技大学 | Moving target traversal access sequence planning method based on improved pointer network |
| CN118394109A (en) * | 2024-06-26 | 2024-07-26 | 烟台中飞海装科技有限公司 | Simulated countermeasure training method based on multi-agent reinforcement learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114815834A (en) | Dynamic path planning method for mobile intelligent agent in stage environment | |
| Henry et al. | Learning to navigate through crowded environments | |
| CN112799386B (en) | Robot path planning method based on artificial potential field and reinforcement learning | |
| Sun et al. | Crowd navigation in an unknown and dynamic environment based on deep reinforcement learning | |
| Ronecker et al. | Deep Q-network based decision making for autonomous driving | |
| CN112132263A (en) | Multi-agent autonomous navigation method based on reinforcement learning | |
| Merabti et al. | Nonlinear predictive control of a mobile robot: a solution using metaheuristcs | |
| Liu et al. | Deep structured reactive planning | |
| Zhang et al. | A unified framework integrating decision making and trajectory planning based on spatio-temporal voxels for highway autonomous driving | |
| Rupprecht et al. | A survey for deep reinforcement learning in markovian cyber–physical systems: Common problems and solutions | |
| CN114898293A (en) | Pedestrian crossing group multi-mode trajectory prediction method for automatically driving automobile | |
| CN116050245A (en) | Highway automatic driving commercial vehicle track prediction and decision method and system based on complex network theory | |
| Wang et al. | Research on dynamic path planning of wheeled robot based on deep reinforcement learning on the slope ground | |
| CN114396949A (en) | Mobile robot no-priori map navigation decision-making method based on DDPG | |
| Chang et al. | Interpretable fuzzy logic control for multirobot coordination in a cluttered environment | |
| Duffy et al. | Agent chameleons: Agent minds and bodies | |
| Thomas et al. | Learning spatiotemporal occupancy grid maps for lifelong navigation in dynamic scenes | |
| Liu et al. | Multi-agent trajectory prediction with graph attention isomorphism neural network | |
| Zhang et al. | Direction-decision learning based pedestrian flow behavior investigation | |
| CN115097814A (en) | Mobile robot path planning method, system and application based on improved PSO algorithm | |
| Li et al. | Predictive hierarchical reinforcement learning for path-efficient mapless navigation with moving target | |
| Zhou et al. | Sliding sequential CVAE with time variant socially-aware rethinking for trajectory prediction | |
| CN112721948A (en) | Method for realizing lane change scheduling of automatic driving automobile based on prediction and search framework | |
| Xie et al. | Hybrid AI-based Dynamic Re-routing Method for Dense Low-Altitude Air Traffic Operations | |
| CN117371895A (en) | Multi-ground unmanned vehicle path planning method, system and medium in unknown environment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |