Intention-Aware Navigation in Crowds with Extended-Space POMDP Planning
Himanshu Gupta
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9445879/Screenshot_from_2022-03-28_18-25-54.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9498915/engineering_and_applied_sciences.png)
Bradley Hayes
Zachary Sunberg
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9502076/zachary_sunberg.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/2125233/images/9503025/cairo-logo-clear-text.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/2125233/images/9503030/headshot.jpg)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9520169/IMG_20220322_194846_153.jpg)
Autonomous systems in the real world
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9470374/213028_New_Volvo_XC40_Pilot_Assist_0.jpg)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9470379/Volkswagen-Golf-production.jpg)
A capable robot must
Infer pedestrian's intention
Predict pedestrian's behavior given its intention
Plan a path
to its goal location
Prior Work
Reactive Controller
Predict and Act Controller
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9478588/temp.drawio.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9478591/temp.drawio__1_.png)
Issues
- Pedestrian Model?
- Future effect of immediate actions?
Issue
- Uncertainty in pedestrian intention estimation?
Need a method that determines the optimal action for the vehicle at any given time by reasoning over the uncertainty in pedestrian intention estimation?
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9502332/superman_pomdp1.jpg)
Need a method that determines the optimal action for the vehicle at any given time by reasoning over the uncertainty in pedestrian intention estimation?
Two-step Approach
Bai et. al, ICRA 2015
Two-step Approach
Bai et. al, ICRA 2015
Solving POMDP using DESPOT
- STATE:
\((x_c,y_c,\theta_c,v_c, g_c)\)
corresponding to the 2D pose, speed and goal of the vehicle.
\((x_i,y_i,v_i, g_i)\)
corresponding to the \(i^{th}\) pedestrian's state
- ACTION:
$$\delta_s \in \{\textbf{Increase Speed, Decrease Speed,}$$ $$\textbf{Maintain Speed, Sudden Brake\}}$$
Effective
Roll-out Policy is important
Bai et. al, ICRA 2015
Two-step Approach
$$\delta_s \in \{\textbf{Increase Speed, Decrease Speed,}$$ $$\textbf{Maintain Speed, Sudden Brake\}}$$
Bai et. al, ICRA 2015
\(\textbf{1D-A}^*\) Approach
ISSUES?
- Decoupling of heading angle planning and speed planning often leads to unnecessary stalling of the vehicle!
- Hybrid A* path can't be found at at every time step!
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9466683/resusing_old_hybrid_astar_path_1D_action_space_speed_pomdp_planner_run.gif)
Bai et. al, ICRA 2015
2D Approach
2D Approach
Solving POMDP using DESPOT
- STATE:
\((x_c,y_c,\theta_c,v_c, g_c)\)
corresponding to the 2D pose, speed and goal of the vehicle.
\((x_i,y_i,v_i, g_i)\)
corresponding to the \(i^{th}\) pedestrian's state
- ACTION:
$$\delta_s(t) \in \{\textbf{Increase Speed, Decrease Speed,}$$ $$\textbf{Maintain Speed, Sudden Brake\}}$$
Same as previous POMDP
- ACTION:
$$\mathcal{a = ( \delta_\theta , \delta_s )}$$
2D Approach
- Critical Challenge: Determining a good roll-out policy for the vastly increased set of states reachable in the tree search.
Effective roll-out policy
- Obtain a path using multi query motion planning technique
- Probabilistic RoadMap (PRM)
- Fast Marching Method (FMM)
- Roll-out policy: Execute a reactive controller over the obtained path
Probabilistic RoadMaps (PRM) for Multi-Query Path Planning
Simulation Environment
- Environment: \(100\)m x \(100\)m square field
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9474135/random_drawings.drawio.png)
-
Autonomous vehicle: A holonomic vehicle.
- Inspired by Kinova MOVO
- Max speed: \(2\) m/s
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9474181/movo_screen_grab.jpg)
Experimental Scenarios
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9466275/scenario1_humans400.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9466276/scenario2_humans400.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9466279/scenario3_humans400.png)
Scenario 1
(Open Field)
Scenario 2 (Cafeteria Setting)
Scenario 3
(L shaped lobby)
Planners
# possible actions in POMDP Planning: 4
# possible actions in POMDP Planning: 11
Experimental Details
- For each scenario, we ran sets of 100 different experiments with different pedestrian densities in the environment.
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9474694/sce1_humans_100.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9474693/sce1_humans_300.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9474696/sce1_humans_400.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9474695/sce1_humans_200.png)
# humans = 100
# humans = 200
# humans = 300
# humans = 400
Scenario 1
Results
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9466242/Screenshot_from_2022-04-05_10-12-21.png)
Evaluation Metric: Travel Time (in s)
Results
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9466242/Screenshot_from_2022-04-05_10-12-21.png)
Evaluation Metric: #Outperformed
Scenario 2
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9466269/sce2_astar.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9466271/sce2_fmm.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9466272/sce2_prm.png)
\(1D-A^*\)
\(2D-FMM\)
\(2D-PRM\)
Results
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9466761/safe_planner2.gif)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9466683/resusing_old_hybrid_astar_path_1D_action_space_speed_pomdp_planner_run.gif)
Limited Space Planner
Extended Space Planner
Conclusion
Thank You!
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9445879/Screenshot_from_2022-03-28_18-25-54.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9498915/engineering_and_applied_sciences.png)
Himanshu Gupta
himanshu.gupta@colroado.edu
Extended Space POMDP Planning
(AAMAS 2022)
https://github.com/himanshugupta1009/extended_space_navigation_pomdp
![](https://s3.amazonaws.com/media-p.slid.es/uploads/2125233/images/9503025/cairo-logo-clear-text.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9520169/IMG_20220322_194846_153.jpg)
Summary
-
We proposed a new technique for navigation problems with arbitrary state uncertainty by incorporating multi-query path planning techniques for effective tree search using online solvers for a POMDP
- Probabilistic RoadMaps
-
Fast Marching Methods
- We demonstrated that our technique outperforms the current state-of-the-art method significantly in terms of travel time while being just as safe!
Experimental Details
- In simulations, the planning time for the vehicle at each step is 0.5 seconds
Experiments (NHV)
Limited Space
Planner
Extended Space
Planners
\(1D\)-\(A^*\)
\(2D\)-\(NHV\)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9466275/scenario1_humans400.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9479852/dubin.png)
Results (NHV)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9466290/Screenshot_from_2022-04-05_10-23-42.png)
Evaluation Metric: Travel Time (in s)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9466290/Screenshot_from_2022-04-05_10-23-42.png)
Evaluation Metric: #Outperformed
Results (NHV)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9466278/nhv_astar.png)
![](https://s3.amazonaws.com/media-p.slid.es/uploads/1988245/images/9466280/nhv_sl.png)
\(1D-A^*\)
\(2D-NHV\)
Results (NHV)
AAMAS Conference Presentation
By Himanshu Gupta
AAMAS Conference Presentation
- 288