Safe and effective navigation of autonomous robots in hazardous environments

Abstract

The development of autonomous mobile machines to perform useful tasks in real work environments is currently being impeded by concerns over effectiveness, commercial viability and, above all, safety. This paper introduces a case study of a robotic excavator to explore a series of issues around system development, navigation in unstructured environments, autonomous decision making and changing the behaviour of autonomous machines to suit the prevailing demands of users. The adoption of the Real-Time Control Systems (RCS) architecture (Albus, 1991) is proposed as a universal framework for the development of intelligent systems. In addition it is explained how the use of Partially Observable Markov Decision Processes (POMDP) (Kaelbling et al., 1998) can form the basis of decision making in the face of uncertainty and how the technique can be effectively incorporated into the RCS architecture. Particular emphasis is placed on ensuring that the resulting behaviour is both task effective and adequately safe, and it is recognised that these two objectives may be in opposition and that the desired relative balance between them may change. The concept of an autonomous system having "values" is introduced through the use of utility theory. Limited simulation results of experiments are reported which demonstrate that these techniques can create intelligent systems capable of modifying their behaviour to exhibit either 'safety conscious' or 'task achieving' personalities.

On Provably Safe Obstacle Avoidance for Autonomous Robotic Ground Vehicles

“…Seward et al [24] try to avoid potentially hazardous situations by using Partially Observable Markov Decision Processes. Their focus, however, is on a user-definable tradeoff between safety and progress towards a goal.…”

“…Unlike existing work on obstacle avoidance (e. g., [1,17,[24][25][26]), we use hybrid models and verification techniques that describe and verify the robot's discrete control choices along with its continuous, physical motion. In summary, our contributions are (i) hybrid models of navigation and obstacle avoidance control algorithms of robots, and (ii) proofs that they guarantee passive and passive friendly safety in the presence of stationary and moving obstacles despite sensor and actuator uncertainty.…”

Safety of Autonomous Cognitive-oriented Robots

“…A more general approach toward autonomous operation of mobile robots is pursued by Seward et al (2000Seward et al ( , 2007. Basically, they develop the robotic excavator LUCIE which is intended to act autonomously.…”

“…However, autonomous systems basically offer the chance, as outlined in Section 3, to adjust their operation with regard to various aspects, including safety. Seward et al (2007) call this the 'self-safety management'. Their central approach to the safety problem is that the self-safety management has to be become an inherent part of the system behavior.…”

Action planning for autonomous systems with respect to safety aspects

“…In [12] and [15] is described a method how a probabilistic model can be defined to enable decision-making under considering of safety aspects. The hazard analysis-realized by a fault tree analysis-is the basis of the quantitative and probabilistic description of accident or hazardous events.…”