Raghvendra V. Cowlagi

Highlights from the literature on accident causation and system safety: Review of major ideas, recent contributions, and challenges

Abstract This work constitutes a short guide to the extensive but fragmented literature on accident causation and system safety. After briefly motivating the interest in accident causation and discussing the notion of a safety value chain, we delve into our multi-disciplinary review with discussions of Man Made Disasters, Normal Accident, and the High Reliability Organizations (HRO) paradigm. The HRO literature intersects an extensive literature on safety culture, a subject we then briefly touch upon. Following this discussion, we note that while these social and organizational contributions have significantly enriched our understanding of accident causation and system safety, they have important deficiencies and are lacking in their understanding of technical and design drivers of system safety and accident causation. These missing ingredients, we argue, were provided in part by the development of Probabilistic Risk Assessment (PRA). The idea of anticipating possible accident scenarios, based on the system design and configuration, as well as its technical and operational characteristics, constitutes an important contribution of PRA, which builds on and extends earlier contributions made by the development of Fault Tree and Event Tree Analysis. We follow the discussion of PRA with an exposition of the concept of safety barriers and the principle of defense-in-depth, both of which emphasize the functions and “safety elements [that should be] deliberately inserted” along potential accident trajectories to prevent, contain, or mitigate accidents. Finally, we discuss two ideas that are emerging as foundational in the literature on system safety and accident causation, namely that system safety is a “control problem”, and that it requires a “system theoretic” approach to be dealt with. We clarify these characterizations and indicate research opportunities to be pursued along these directions. We conclude this work with two general recommendations: (1) that more fundamental research and cross-talk across several academic disciplines must be supported and incentivized for tackling the multi-disciplinary issues of accident causation and system safety (e.g., through the creation “academic hubs” or “centers of excellence” dedicated to system safety); and (2) that more interactions and partnerships between academia, industry, and government (especially accident investigation agencies) on accident causation and system safety issues would be particularly useful for all involved in advancing the safety agenda, from both research and education perspectives, and for disseminating research results, safety recommendations, and lessons learned from accident investigations.

Optimal motion planning with the half-car dynamical model for autonomous high-speed driving

We discuss an implementation of the RRT* optimal motion planning algorithm for the half-car dynamical model to enable autonomous high-speed driving. To develop fast solutions of the associated local steering problem, we observe that the motion of a special point (namely, the front center of oscillation) can be modeled as a double integrator augmented with fictitious inputs. We first map the constraints on tire friction forces to constraints on these augmented inputs, which provides instantaneous, state-dependent bounds on the curvature of geometric paths feasibly traversable by the front center of oscillation. Next, we map the vehicles actual inputs to the augmented inputs. The local steering problem for the half-car dynamical model can then be transformed to a simpler steering problem for the front center of oscillation, which we solve efficiently by first constructing a curvature-bounded geometric path and then imposing a suitable speed profile on this geometric path. Finally, we demonstrate the efficacy of the proposed motion planner via numerical simulation results.

Multiresolution path planning with wavelets: A local replanning approach

A path planning algorithm based on multiresolution cell decomposition of the environment using wavelets is proposed. The environment is assumed to be given by an occupancy grid at fine resolution. The algorithm constructs a cell decomposition at several levels of resolution (cell sizes) and constructs an optimal path to the destination from the current location of the agent. At each step the algorithm iteratively refines a coarse approximation to the path through local replanning. The replanning process uses previous information to refine the original cell channel in the immediate area of the path. This is done efficiently using the wavelet coefficients. Numerical tests show a speed-up of an order of magnitude over the baseline algorithm with minimal impact on the overall optimality of the resulting path. A comparative study with the well-known D* algorithm is also provided.

Coordinability and consistency in accident causation and prevention: formal system theoretic concepts for safety in multilevel systems

Although a system approach to accidents in sociotechnical systems has been frequently advocated, formal system theoretic concepts remain absent in the literature on accident analysis and system safety. To address this gap, we introduce the notions of coordinability and consistency from the hierarchical and multilevel systems theory literature. We then investigate the applicability and the importance of these concepts to accident causation and safety. Using illustrative examples, including the worst disaster in aviation history, and recent incidents in the United States of aircraft clipping each other on the tarmac, we propose that the lack of coordinability is a fundamental failure mechanism causing or contributing to accidents in multilevel systems. We make a similar case for the lack of consistency. Coordinability and consistency become ingredients for accident prevention, and their absence fundamental failure mechanisms that can lead to system accidents. Finally, using the concepts introduced in this work, we identify several venues for further research, including the development of a theory of coordination in multilevel systems, the investigation of potential synergies between coordinability, consistency, and the high reliability organizations paradigm, and the possibility of reframing the view that sloppy management is the root cause of many industrial accidents as one of lack of coordinability and/or consistency between management and operations. By introducing and expanding on the concepts of coordinability and consistency, we hope to contribute to the thinking about, and the to language of, accident causation, and prevention and to add to the intellectual toolkit of safety professionals and academics.

Beyond quadtrees: Cell decompositions for path planning using wavelet transforms

Path planning techniques based on hierarchical multiresolution cell decompositions are suitable for online implementation due to their simplicity and speed of implementation. We present an efficient multiresolution cell decomposition scheme based on the Haar wavelet transform. The decomposition approximates the environment using high resolution close to the agent and coarse resolution elsewhere. We demonstrate an algorithm to extract the adjacency and transition cost relations of the cells directly from the wavelet transform coefficients.

Shortest distance problems in graphs using history-dependent transition costs with application to kinodynamic path planning

A new algorithm is presented to compute the shortest path on a graph when the node transition costs depend on the prior history of the path to the current node. The algorithm is applied to solve path planning problems with curvature constraints.

An Offline/Online DDDAS Capability for Self-Aware Aerospace Vehicles

In this paper we develop initial offline and online capabilities for a self-aware aerospace vehicle. Such a vehicle can dynamically adapt the way it performs missions by gathering information about itself and its surroundings via sensors and responding intelligently. The key challenge to enabling such a self-aware aerospace vehicle is to achieve tasks of dynamically and autonomously sensing, planning, and acting in real time. Our first steps towards achieving this goal are presented here, where we consider the execution of online mapping strategies from sensed data to expected vehicle capability while accounting for uncertainty. Libraries of strain, capability, and maneuever loading are generated offline using vehicle and mission modeling capabilities we have developed in this work. These libraries are used dynamically online as part of a Bayesian classification process for estimating the capability state of the vehicle. Failure probabilities are then computed online for specific maneuvers. We demonstrate our models and methodology on decisions surrounding a standard rate turn maneuver.

Multiresolution Motion Planning for Autonomous Agents via Wavelet-Based Cell Decompositions

We present a path- and motion-planning scheme that is “multiresolution” both in the sense of representing the environment with high accuracy only locally and in the sense of addressing the vehicle kinematic and dynamic constraints only locally. The proposed scheme uses rectangular multiresolution cell decompositions, efficiently generated using the wavelet transform. The wavelet transform is widely used in signal and image processing, with emerging applications in autonomous sensing and perception systems. The proposed motion planner enables the simultaneous use of the wavelet transform in both the perception and in the motion-planning layers of vehicle autonomy, thus potentially reducing online computations. We rigorously prove the completeness of the proposed path-planning scheme, and we provide numerical simulation results to illustrate its efficacy.

Curvature-Bounded Traversability Analysis in Motion Planning for Mobile Robots

We consider the geometric problem of deciding whether a narrow planar passage can be traversed by a curve that satisfies prespecified upper bounds on its curvature. This problem is of importance for path- and motion-planning of autonomous mobile robots, particularly when vehicle dynamical constraints are considered during planning. For a special case of narrow passages, namely, rectangular channels, we present a fast numerical algorithm to determine if a given channel may be traversed via curvature-bounded paths. We demonstrate that the proposed algorithm can affirm traversability in cases where the most recent result in the literature fails.

Multi-resolution path planning: Theoretical analysis, efficient implementation, and extensions to dynamic environments

A multi-resolution path planning algorithm based on the wavelet transform of the environment has been reported previously in the literature. In this paper, we provide a proof of completeness of this algorithm. In addition, we present an implementation of this algorithm that reuses information obtained in previous iterations to perform subsequent iterations more efficiently. Finally, we extend this path planning algorithm to dynamic environments by presenting a simple scheme for updating the wavelet transform coefficients to reflect changes in the environment.