Uncertainty-based Online Mapping and Motion Planning for Marine Robotics Guidance
read more
Citations
Online motion planning for unexplored underwater environments using autonomous underwater vehicles
Online Mapping and Motion Planning under Uncertainty for Safe Navigation in Unknown Environments
A Decentralised Strategy for Heterogeneous AUV Missions via Goal Distribution and Temporal Planning
Towards robust grasps: Using the environment semantics for robotic object affordances
Learning and Generalisation of Primitives Skills Towards Robust Dual-arm Manipulation
References
Probabilistic roadmaps for path planning in high-dimensional configuration spaces
Randomized kinodynamic planning
OctoMap: an efficient probabilistic 3D mapping framework based on octrees
The Open Motion Planning Library
Randomized kinodynamic planning
Related Papers (5)
Frequently Asked Questions (13)
Q2. What is the effect of the local map LMn?
The local map LMn expands the environment awareness with the occluded (hidden) regions, i.e. unseen voxels that are within the maximum sensor’s range but located behind the obstacles (occupied space).
Q3. How was the robot able to navigate in the environment?
In both simulated and in-water experiments, the robot was required to map the environment with a 0.5m map resolution, and to navigate at a constant depth of 1.5m with a maximum velocity of 0.35m/s.
Q4. What are the main requirements of the planning problem?
The planning problem defined in Section II-D has three main requirements: to consider the vehicle’s motion capabilities, to validate probabilistic constraints, and to meet online computation limitations.
Q5. Why does the planner have a high success rate when solving start-to-goal queries?
This performance is because the planner exploits the entire system’s dynamic range, thus allowing the vehicle to execute more complex manoeuvres by adjusting the surge (forward) velocity and prioritising the turning rate.
Q6. How many times did the robot solve the start-to-goal query?
The robot was required to solve a start-to-goal-query to reach a goal region Bgoal located on the opposite side of the structure, which can only be achieved by navigating through any of the narrow four-metre gaps.
Q7. What is the higher success rate with respect to the previous experiment?
The higher success rate with respect to the previous experiment is given by the nature of the environment; this scenario involves less abrupt manoeuvres and the passage is wider, more than twice larger though.
Q8. What is the probability of a robot being in collision with an obstacle?
More specifically, the probability of the system being in collision with an obstacle in the environment at time k is characterised as:pcollision(bk,M) ≤ ∫X bk(x)FO(x) dx=∫X N (x | ẑk, Pzk)FO(x) dx. (7)Given a minimum probability of safety psafe, the authors require 1− pcollision(b, M) > psafe for every belief b on the path in order to probabilistically guarantee the robot’s safety.
Q9. What is the relative uncertainty between the two elements?
Retrieving the voxel’s timestamp ki and the timestamp kj of the state ẑk, the relative uncertainty between these elements is defined as:Prelative =kj−1∑m=kiAkj−m−1
Q10. What is the alternative to the shooting approach?
aiming to preserve the entire system’s dynamic range, their planning module uses a shooting approach, which consists of expanding the tree from the node with the lowestcost within a neighbourhood of δ-radius.
Q11. How was the framework evaluated in the simulated environment?
The minimum probability of safeness for all experiments was set at psafe = 0.9.1) Simulated trials: before conducting in-water experiments, the framework was exhaustively tested in the simulated breakwater structure and canyon scenarios.
Q12. What is the main difference between the two approaches?
On top of that, and in contrast to the approaches in [12] and [13], the proposed kernelbased method deals with nonconvex representations of the environment, thus avoiding convexification routines, which usually involve a loss of the environment awareness and an increase of the computation time.
Q13. What is the title of the paper?
1A complete sea-trial through the real breakwater structure can be seen in: https://youtu.be/dTejsNqNC00.This paper has presented a novel end-to-end framework which probabilistically guarantees the robot’s safety when solving start-to-goal queries in unexplored environments.