Time scaled collision cone based trajectory optimization approach for reactive planning in dynamic environments

Gopalakrishnan, Bharath; Singh, Arun Kumar; Krishna, K. Madhava

doi:10.1109/iros.2014.6943150

Cited by 23 publications

(17 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…15,16 In the latter case, a simple cost function is generally used, for example, minimal distance to the goal. It is important to emphasize that there is a rich literature in collision avoidance approaches; 6,7,[17][18][19][20] however, all of them provide only collision-free guarantee where motion safety is not considered at all. Motion safety requires guaranteeing that collision will never occur.…”

Section: Related Workmentioning

confidence: 99%

Safe Motion Planning Based on a New Encoding Technique for Tree Expansion Using Particle Swarm Optimization

Bouraine

Azouaoui

2020

Robotica

View full text Add to dashboard Cite

SUMMARY Robots are now among us and even though they compete with human beings in terms of performance and efficiency, they still fail to meet the challenge of performing a task optimally while providing strict motion safety guarantees. It is therefore necessary that the future generation of robots evolves in this direction. Generally, in robotics state-of-the-art approaches, the trajectory optimization and the motion safety issues have been addressed separately. An important contribution of this paper is to propose a motion planning method intended to simultaneously solve these two problems in a formal way. This motion planner is dubbed PassPMP-PSO. It is based on a periodic process that interleaves planning and execution for a regular update of the environment’s information. At each cycle, PassPMP-PSO computes a safe near-optimal partial trajectory using a new tree encoding technique based on particle swarm optimization (PSO). The performances of the proposed approach are firstly highlighted in simulation environments in the presence of moving objects that travel at high speed with arbitrary trajectories, while dealing with sensors field-of-view limits and occlusions. The PassPMP-PSO algorithm is tested for different tree expansions going from 13 to more than 200 nodes. The results show that for a population between 20 and 100 particles, the frequency of obtaining optimal trajectory is 100% with a rapid convergence of the algorithm to this solution. Furthermore, an experiment-based comparison demonstrates the performances of PassPMP-PSO over two other motion planning methods (the PassPMP, a previous variant of PassPMP-PSO, and the input space sampling). Finally, PassPMP-PSO algorithm is assessed through experimental tests performed on a real robotic platform using robot operating system in order to confirm simulation results and to prove its efficiency in real experiments.

show abstract

Section: Related Workmentioning

confidence: 99%

Safe Motion Planning Based on a New Encoding Technique for Tree Expansion Using Particle Swarm Optimization

Bouraine

Azouaoui

2020

Robotica

View full text Add to dashboard Cite

show abstract

“…In this section, we show that velocity optimization (13a)-(13f) can be reformulated to a convex quadratic programming (QP) problem. We start with the convexification of TSCC constraints (10) or (12) and then subsequently show that its intersection space with velocity and acceleration bounds can be described in terms of linear inequalities. For the ease of exposition, we introduce the following change of variables [23].ṡ…”

Section: Simplification Of Velocity Optimizationmentioning

confidence: 99%

Model Predictive Control for Autonomous Driving Based on Time Scaled Collision Cone

Babu

Oza

Singh

et al. 2018

2018 European Control Conference (ECC)

Self Cite

View full text Add to dashboard Cite

In this paper, we present a Model Predictive Control (MPC) framework based on path velocity decomposition paradigm for autonomous driving. The optimization underlying the MPC has a two layer structure wherein first, an appropriate path is computed for the vehicle followed by the computation of optimal forward velocity along it. The very nature of the proposed path velocity decomposition allows for seamless compatibility between the two layers of the optimization.A key feature of the proposed work is that it offloads most of the responsibility of collision avoidance to velocity optimization layer for which computationally efficient formulations can be derived. In particular, we extend our previously developed concept of time scaled collision cone (TSCC) constraints and formulate the forward velocity optimization layer as a convex quadratic programming problem. We perform validation on autonomous driving scenarios wherein proposed MPC repeatedly solves both the optimization layers in receding horizon manner to compute lane change, overtaking and merging maneuvers among multiple dynamic obstacles.

show abstract

“…In order to show the robustness of our approach over different trajectory generators, we combine our RL based approach with Bernstein curves [7] for trajectory generation and with the ROS Navigation Stack, in which navigation carried out in two levels where Informed RRT* [6] behaves as the global planner and Timed Elastic Band (TEB) [21] performs the role of the trajectory generator. We show results both without (naive) and with RL, to show that an informed approach using RL is more effective than an uninformed or naive path planner.…”

Section: Goal Based Navigationmentioning

confidence: 99%

“…By using RL in an episodic manner, the result of each failure is propagated back to earlier states over multiple episodes, making the predictions of /submit/2514950/addfilesfuture failures more likely in the earlier stages once the learning converges. Analysis shows significant reduction in SLAM failures when the learned Q values are used to filter actions in trajectories generated by routine methods such as sampling based planners [13] or trajectory optimization routines [7] [21]. Qualitative and quantitative analysis and comparisons with supervised learning approaches, 3D point overlap maximization techniques, and predicting Localization Quality estimates, described in [16], showcase the superior performance of the proposed method.…”

Section: Introductionmentioning

confidence: 99%

Learning to Prevent Monocular SLAM Failure using Reinforcement Learning

Prasad¹,

Yadav

Saurabh

et al. 2018

Proceedings of the 11th Indian Conference on Computer Vision, Graphics and Image Processing

View full text Add to dashboard Cite

Monocular SLAM refers to using a single camera to estimate robot ego motion while building a map of the environment. While Monocular SLAM is a well studied problem, automating Monocular SLAM by integrating it with trajectory planning frameworks is particularly challenging. This paper presents a novel formulation based on Reinforcement Learning (RL) that generates fail safe trajectories wherein the SLAM generated outputs do not deviate largely from their true values. Quintessentially, the RL framework successfully learns the otherwise complex relation between perceptual inputs and motor actions and uses this knowledge to generate trajectories that do not cause failure of SLAM. We show systematically in simulations how the quality of the SLAM dramatically improves when trajectories are computed using RL. Our method scales effectively across Monocular SLAM frameworks in both simulation and in real world experiments with a mobile robot.

show abstract

Time scaled collision cone based trajectory optimization approach for reactive planning in dynamic environments

Cited by 23 publications

References 16 publications

Safe Motion Planning Based on a New Encoding Technique for Tree Expansion Using Particle Swarm Optimization

Safe Motion Planning Based on a New Encoding Technique for Tree Expansion Using Particle Swarm Optimization

Model Predictive Control for Autonomous Driving Based on Time Scaled Collision Cone

Learning to Prevent Monocular SLAM Failure using Reinforcement Learning

Contact Info

Product

Resources

About