Passively safe partial motion planning for mobile robots with limited field-of-views in unknown dynamic environments

Abstract

Abstract-This paper addresses the problem of planning the motion of a mobile robot with a limited sensory field-of-view in an unknown dynamic environment. In such a situation, the upper-bounded planning time prevents from computing a complete motion to the goal, partial motion planning is in order. Besides the presence of moving obstacles whose future behaviour is unknown precludes absolute motion safety (in the sense that no collision will ever take place whatever happens) is impossible to guarantee. The stance taken herein is to settle for a weaker level of motion safety called passive motion safety: it guarantees that, if a collision takes place, the robot will be at rest. The primary contribution of this paper is PASSPMP, a partial motion planner enforcing passive motion safety. PASSPMP periodically computes a passively safe partial trajectory designed to drive the robot towards its goal state. Passive motion safety is handled using a variant of the Inevitable Collision State (ICS) concept called Braking ICS, i.e. states such that, whatever the future braking trajectory of the robot, a collision occurs before it is at rest. Simulation results demonstrate how PASSPMP operates and handles limited sensory field-of-views, occlusions and moving obstacles with unknown future behaviour. More importantly, PASSPMP is provably passively safe.

Tackling Occlusions & Limited Sensor Range with Set-based Safety Verification

“…Some of them explicitly consider uncertainties [4], [7], [11], while others use visibility analysis to define velocity constraints [5], [6]. Only two publications among those prove at least passive motion safety 2 [10] or prove collision freedom at discrete time steps of their trajectories 3 [11].…”

“…"If a collision takes place, the robot will be at rest. "[10] 3 Collisions between discrete time steps will not be detected here 4. Meaning higher than passive or even passive friendly safety[12], but at least RSS[1].…”

Perception, Planning, Control, and Coordination for Autonomous Vehicles

“…In [242], a finite state machine generates new sub-goals for a sampling based replanner only when the robot was forced to come to a stop due to path blockage, where each subgoal was set at a predefined distance ahead along the predefined path. In [254], no sub-goals were defined, nor was there a predefined path to utilize, but rather the best choice trajectories were decided based on a combined weighted heuristic of trajectory execution time and distance to goal from the end trajectory state. Bouraine et al [254] applied a constant rate replanning timer, where each current solution plan was executed concurrently while the subsequent plan was being generated, and each newly planned trajectory would be rooted from an anticipated committed pose given the previous committed solution trajectory.…”

“…In [254], no sub-goals were defined, nor was there a predefined path to utilize, but rather the best choice trajectories were decided based on a combined weighted heuristic of trajectory execution time and distance to goal from the end trajectory state. Bouraine et al [254] applied a constant rate replanning timer, where each current solution plan was executed concurrently while the subsequent plan was being generated, and each newly planned trajectory would be rooted from an anticipated committed pose given the previous committed solution trajectory. Note that in a Mobility-on-Demand (MoD) context, a mission planner should be able to provide a predefined path which leads from a starting point to an end destination based on a passenger service request, and the presence of a predefined path can help to overcome dangers of getting stuck due to local minima.…”

Model-based robocentric planning and navigation for dynamic environments

“…Bouraine et al (2012) defined Braking-ICS , a version of ICS that guarantees that if a collision takes place, the robot will be at rest. It was later applied to the development of a partial motion planner by Bouraine et al (2014).…”