Inevitable Collision States Research Articles

Handling inevitable collision states by Advanced Driver Assistance Systems functions: software-in-the-loop performance assessment of an injury risk-based logic in a “lane departure” scenario

The downward trend in the number of fatalities and serious injuries related to road accidents depends on the implementation of increasingly performing Advanced Driver Assistance Systems (ADAS) in the circulating fleet. The greatest benefit of the adoption of ADASs like Autonomous Emergency Braking (AEB) consists in limiting the frequency of impacts. However, in Inevitable Collision States (ICSs), the decrease in impact closing speed guaranteed by the AEB may not reduce the Injury Risk (IR) for the occupants: IR is a function of the vehicle’s velocity change in the collision () – a combination of impact closing speed and impact eccentricity. The work virtually analyses, in lane departure ICS scenarios, the performance of an adaptive steering and braking intervention logic based on instantaneous IR minimization. The adaptive logic reduces IR compared to the absence of intervention (down to 80 times lower) and to the AEB (down to 40 times lower) by leading the ego vehicle toward eccentric impact configurations. It is highlighted that full activation of the steer-by-wire system in 0.3 s allows the adaptive logic to also reduce the frequency of impacts; it is further evidenced that employing a function capable of modulating the braking level to minimize IR entails disadvantages from the IR perspective compared to the AEB: efficient intervention strategies on the steering are the only alternative for increasing the safety provided by high-performance ADASs. Finally, compared to previous literature, the study highlights high efficiencies of the adaptive logic in a wide range of ICS scenarios.

Performance Assessment in a “Lane Departure” Scenario of Impending Collision for an ADAS Logic Based on Injury Risk Minimisation

The current prioritisation of road safety enhancement in the automotive sector is leading toward the near future implementation of Advanced Driver Assistance Systems (ADASs), aiming at the simultaneous intervention of braking and steering for impact avoidance in case of an impending collision. However, it is partially unclear how new technologies for controlling the steering will actually behave in the case of inevitable collision states; the need consequently emerges to propose and tune efficient ADAS strategies to handle the complexity of critical road scenarios. An adaptive intervention logic on braking and steering for highly automated vehicles is applied in the context of a “lane departure”, two-vehicle critical road scenario; the ADAS implementing the logic activates to minimise the injury risk for the ego vehicle’s occupants at each time step, adapting to the eventual scenario evolution consequent to actions by other road users. The performance of the adaptive logic is investigated by a software-in-the-loop approach, varying the mutual position of the involved vehicles at the beginning of the criticality and comparing the injury risk outcomes of the eventual impacts with those connected to the Autonomous Emergency Braking (AEB). The results highlight a twofold benefit from the adaptive logic application in terms of road safety: (1) it decreases the frequency of impacts compared to the AEB function; (2) in inevitable collision states, it decreases injury risk for the vehicles’ occupants down to 40% compared to the AEB. This latter condition is achieved thanks to the possibility of reaching highly eccentric impact conditions (low impact forces and occupants’ injury risk as a consequence). The obtained highlights expand the literature regarding the adaptive logic by considering a diverse critical road scenario and investigating how fine variations on the vehicles’ mutual position at the beginning of the criticality reflect on the injury outcomes for different types of intervention logic.

Combined activation of braking and steering for automated driving systems: adaptive intervention by injury risk-based criteria

Increase in advanced driver assistance systems (ADAS) performances is a crucial step towards autonomous driving, allowing the design of increasingly reliable automated driving systems (ADS); ADAS devices play a key role in the enhancement of vehicle safety, which primarily results from the ability to avoid possible impacts. Nevertheless, inevitable collision states (ICS) can be triggered by obstacles as buildings and stationary vehicles interposing between the opponent and the working field of ADAS sensors, compromising their functions; therefore, the performance increase of ADAS devices on the market necessarily passes from the optimal handling of an ICS, which is not currently subject to evaluations.The work introduces ADAS intervention criteria which are based on the occupants’ injury risk (IR): in a specific road scenario, the ADAS must primarily avoid the collision with maximum margin and minimize IR in the case of an ICS. Specifically, the ADAS must monitor the environment and intervene on braking and steering adapting to the scenario evolution, following an ”adaptive” logic. The most critical aspect of the approach lies in reconstructing, for the specific intervention, the eventual impact: while being a time-consuming process, reconstruction of the impact phase is necessary to compute impact-related parameters (e.g., velocity change of the vehicle ∆V) which directly affect IR.To highlight the benefits offered by an adaptive ADAS compared to traditional ADASs, a special testing software has been developed: the best adaptive intervention to be applied at each instant is chosen in real-time through the criteria proposed, retrieving the required information from a pre-calculated database which collects the results of each braking and steering manoeuvre for a large variety of critical scenarios. Analyzing three ICS conditions, it is shown that the adaptive logic, differing from an autonomous emergency braking, aims at creating eccentrical impacts with minimum ∆V: the IR values associated with the ADAS adaptive intervention are consequently an order of magnitude lower than those obtained through traditional ADAS interventions.

COLREG-RRT: An RRT-Based COLREGS-Compliant Motion Planner for Surface Vehicle Navigation

Motion planning for autonomous surface vehicles (ASVs) is challenging since surface vessels are nonlinear underactuated kinodynamic systems with often large inertia. Thus, ASV planners must identify long-term trajectories in order to avoid guiding the ASV into inevitable collision states. Furthermore, maritime vessels are required to follow COLlision REGulationS (COLREGS), which dictate collision avoidance patterns. Current state-of-the-art methods are based on model predictive control (MPC) and assume that other vessels move at constant velocities without the consideration of COLREGS. In this letter, we propose COLREG-RRT, an RRT-based planner capable of identifying long-term, COLREGS-compliant trajectories with a high navigation success rate. This is achieved by conducting joint forward simulations of both the ASV and the other vessels during RRT growth in order to anticipate future collisions. The COLREGS-compliance is enforced by constructing virtual obstacles that inhibit tree growth. We demonstrate COLREGS-compliance in single-ship encounters and compare against two state-of-the-art methods in multiship encounters with up to 20 other vessels. Experiments indicate that COLREG-RRT has a 32% higher success rate and is real-time capable in the most difficult environment tested. Additionally, COLREG-RRT identifies longer trajectories, as compared to MPC. This property aids with collision avoidance with other ships.

Inevitable Collision States for Motorcycle-to-Car Collision Scenarios

This paper presents a method to identify inevitable collision states (ICS) specifically for a motorcycle when interacting with an opponent passenger car in typical traffic scenarios. Previous ICS methods were applied to passenger cars or generic vehicles; however, the peculiarities of motorcycles urge the definition of specific methods for these vehicles. The findings extend the applicability of previous algorithms to include all motorcycle-to-car collisions, irrespective of collision configurations. ICS identification can be adopted as a triggering criterion for more invasive safety systems such as motorcycle autonomous emergency braking (MAEB), which require a last-resort approach in their initial phases of development. In this regard, this paper also presents an evaluation of an idealized MAEB through experiments simulating real-world crashes in a computer-based virtual environment.

Comparing Collision Threat Measures for Verification of Autonomous Vehicles using Extreme Value Theory

The verification of safety is expected to be one of the largest challenges in the commercialization of autonomous vehicles. Using traditional methods would require infeasible time and resources. Recent research has shown the possibility of using near-collisions in order to estimate the frequency of actual collisions using Extreme Value Theory. However, little research has been done on how the measure for determining the closeness to a collision affect the result of the estimation. This paper compares a collision-based measure against one that relates to an inevitable collision state. The result shows that using inevitable collision states is more robust and that more research needs to be made into measures of collision proximity.

Application of time dependent probabilistic collision state checkers in highly dynamic environments.

When computing the trajectory of an autonomous vehicle, inevitable collision states must be avoided at all costs, so no harm comes to the device or pedestrians around it. In dynamic environments, considering collisions as binary events is risky and inefficient, as the future position of moving obstacles is unknown. We introduce a time-dependent probabilistic collision state checker system, which traces a safe route with a minimum collision probability for a robot. We apply a sequential Bayesian model to calculate approximate predictions of the movement patterns of the obstacles, and define a time-dependent variation of the Dijkstra algorithm to compute statistically safe trajectories through a crowded area. We prove the efficiency of our methods through experimentation, using a self-guided robotic device.

Provably safe navigation for mobile robots with limited field-of-views in dynamic environments

This paper addresses the problem of navigating in a provably safe manner a mobile robot with a limited field-of-view placed in a unknown dynamic environment. In such a situation, absolute motion safety (in the sense that no collision will ever take place whatever happens in the environment) is impossible to guarantee in general. It is therefore settled for a weaker level of motion safety dubbed passive motion safety: it guarantees that, if a collision takes place, the robot will be at rest. The primary contribution of this paper is the concept of Braking Inevitable Collision States (ICS), i.e. a version of the ICS corresponding to passive motion safety. Braking ICS are defined as states such that, whatever the future braking trajectory followed by the robot, a collision occurs before it is at rest. Passive motion safety is obtained by avoiding Braking ICS at all times. It is shown that Braking ICS verify properties that allow the design of an efficient Braking ICS-Checking algorithm, i.e. an algorithm that determines whether a given state is a Braking ICS or not. To validate the Braking ICS concept and demonstrate its usefulness, the Braking ICS-Checking algorithm is integrated in a reactive navigation scheme called PassAvoid. It is formally established that PassAvoid is provably passively safe in the sense that it is guaranteed that the robot will always stay away from Braking ICS no matter what happens in the environment.

Inevitable collision states — a step towards safer robots?

An inevitable collision state for a robotic system can be defined as a state for which, no matter what the future trajectory followed by the system is, a collision with an obstacle eventually occurs. An inevitable collision state takes into account the dynamics of both the system and the obstacles, fixed or moving. The main contribution of this paper is to lay down and explore this novel concept (and the companion concept of inevitable collision obstacle). Formal definitions of the inevitable collision states and obstacles are given. Properties fundamental for their characterization are established. This concept is very general, and can be useful both for navigation and motion planning purposes (for its own safety, a robotic system should never find itself in an inevitable collision state). To illustrate the interest of this concept, it is applied to a problem of safe motion planning for a robotic system subject to sensing constraints in a partially known environment (i.e. that may contain unexpected obstacles). In safe motion planning, the issue is to compute motions for which it is guaranteed that, no matter what happens at execution time, the robotic system never finds itself in a situation where there is no way for it to avoid collision with an unexpected obstacle.