Dynamic Obstacle Avoidance in uncertain environment combining PVOs and Occupancy Grid
read more
Citations
Reciprocal n-Body Collision Avoidance
Reciprocal Velocity Obstacles for real-time multi-agent navigation
The Hybrid Reciprocal Velocity Obstacle
ClearPath: highly parallel collision avoidance for multi-agent simulation
Safe Maritime Autonomous Navigation With COLREGS, Using Velocity Obstacles
References
Robot Motion Planning
The dynamic window approach to collision avoidance
Using occupancy grids for mobile robot perception and navigation
Motion Planning in Dynamic Environments Using Velocity Obstacles
Heuristic Motion Planning in Dynamic Environments Using Velocity Obstacles
Related Papers (5)
The dynamic window approach to collision avoidance
Frequently Asked Questions (18)
Q2. What future works have the authors mentioned in the paper "Combining probabilistic velocity obstacles and occcupancy grid for safe navigation in dynamic environments" ?
The case of holonome robot and linear constant motion of the obstacles has been analysed in this paper: future work will deal with the generalisation of the method following the Non Linear Velocity Obstacle approach [ 10 ], [ 17 ]. In order to achieve a better performance, the authors plan then to integrate the information of the probability and time to collision in a motion planning algorithm able to face more complex scenarios, combining a priori knowledge and on-line perception, and to test the method on a real mobile base ( Cycab [ 18 ] ).
Q3. What is the definition of the velocity space?
The velocity space is defined as the configuration space where linear velocities are described by vectors attached to the centre of objects.
Q4. What is the probability threshold for the occlusion of the robot?
The velocities that can be represented in the dynamic occupancy grid are integer values in the interval vx = [−4, 4], vy = [−4, 4]; a maximum time of prediction is fixed a T = 5 and the probability threshold is fixed at 0.1.
Q5. What is the advantage of the cell-to-cell approach to the linear velocity obstacles?
The cell-to-cell approach to the linear velocity obstacles allows to reduce the hypothesis of the method, taking into consideration robot and obstacles of whatever shape and whatever discretized approximation of uncertainty in position and velocity of the obstacles.
Q6. What is the input to the algorithm?
The input to the algorithm is an occupancy grid, it is highly reactive to the environmental changes and is well suited to be applied in various sensor settings.
Q7. What is the effect of the robot on the environment?
The robot adapts its behaviour to the quality of information received and modifies its trajectory according to the incomplete and uncertain perception of the environment.
Q8. How does the algorithm generalise to the input provided by an occupancy grid?
In order to generalise the method for a probabilistic approach and to the input provided by an occupancy grid, the authors developed a cell-to-cell approach.
Q9. What is the advantage of considering a dynamic occupancy grid?
The advantage of considering a dynamic occupancy grid is that the robot maintains a full probabilistic information about the present occupation of the space and an estimation of the velocity of each occupied cell in the spatial grid.
Q10. What is the probability of collision in the dynamic and probabilistic case?
In the dynamic and probabilistic case, the navigation of the mobile robot has to attend two major issues: minimise the risk of collision and reach the goal position.
Q11. What is the effect of the robot on the path?
Since the beginning the robot tries to keep its trajectory further away from obstacles; the path results longer as the robot performs wider curves to avoid collisions.
Q12. What is the probability of collision in the case of a robot?
Also in the case of no obstacles in the space, the robot will not move too fast if its perception is limited to a short range or some portion of the space is occluded.
Q13. What is the probability of a collision between a cell and an obstacle?
Since the authors are working with a probabilistic representation, each cell has in general a positive probability of occupation: the free space scanned by the sensor is characterised by a probability of occupation that is nearly 0 while not sensed environment has P (Occ) = 0.5.
Q14. What is the probability of collision for a robot?
This means a velocity v can be applied for an interval ǫ if the robot will not run into collision up to:Tsafe(v) = ǫ + Tbrake(v) (10)where Tbrake(v) is the minimum time to stop applying the maximum negative linear acceleration.
Q15. What is the definition of the robot?
With this definition of the robot, the configuration space ( i.e. the space where each point corresponds to a configuration of the robot and obstacles correspond to configurations in collision) is equivalent to the Euclidean space.
Q16. What is the state of the system x(t) at time t?
The estimation of the state of the system x(t) at time t is the list of the states of all the cells of the grid: Occ, when the cell is occupied or Emp if the correspondent space is free.
Q17. what is the probability of collision of a cell with an obstacle?
1. Considering the whole robot dimension, the maximum probability of collision in the interval [t−1, t] is kept for each object k :Pcoll(vr, k, vn) = max o∈O Po(Occ) · Po(vn) · δ(ko) (6)where o is each cell in SOt(vr′ , r) and δ(ko) = 1 if ko = k, 0 otherwise.
Q18. What is the probability of collision for a given time step?
For each time step, the admissible velocities of the robot are computed taking into account its kinematic and dynamic constraints.