Skip to Main content Skip to Navigation

Lambda-Field : a novel framework for risk assessment in occupancy grids

Abstract : In the context of autonomous robots, one of the most contentious topics is the notion of risk. Indeed, no robot escapes from such a question, that is whether robots will not cause any harm to themselves or the living beings in the surrounding environment. Robotics arms have a finite, relatively small workspace where the risk is tackled in a way that the robot completely stops whenever a human enters its workspace. In mobile robotics, this notion is more complex and still an open problem. First, a representation of the environment is needed for the robot to navigate in it. Oftentimes, the preferred representation of the environment is the semantic one, where each obstacle is stored as a single, unique entity. However, in complex scenarios or unstructured environments, detecting such obstacles is a tedious task and missing one could lead to disastrous events. In these cases, a metric map is used where each position stores the information of occupancy. The most common type of metric map is the Bayesian occupancy map. However, this type of map is not well-fitted to perform risk assessment for continuous paths due to its discrete nature. Hence, we introduce in this thesis a novel type of map called Lambda-Field, specially designed for risk assessment. The Lambda-Fields are a counterpart of the classical Bayesian occupancy grid. Instead of storing the probability of occupancy at each position, the Lambda-Field stores the intensity that a collision will occur at this position: the higher the intensity, the higher the probability of collision. Using this novel formulation, the Lambda-Fields are able to assess a generic risk over a path. Contrary to the Bayesian occupancy grid, the use of intensity instead of directly the probability of collision allows the risk assessment framework to produce physic-based metrics that conserve their physical units. Throughout this thesis, we present how to construct and use the Lambda-Field in both static and dynamic environments. We demonstrate that the Lambda-Field also possesses interesting mapping properties that induce more accurate maps of unstructured environments. Using this risk definition and the Lambda-Field, we show that our framework is capable of doing classical path planning but also cross unstructured environments where a Bayesian occupancy grid would not find any path.
Document type :
Complete list of metadata
Contributor : ABES STAR :  Contact
Submitted on : Tuesday, May 10, 2022 - 2:14:38 PM
Last modification on : Wednesday, May 11, 2022 - 3:48:53 AM


Version validated by the jury (STAR)


  • HAL Id : tel-03663791, version 1


Johann Laconte. Lambda-Field : a novel framework for risk assessment in occupancy grids. Automatic. Université Clermont Auvergne, 2021. English. ⟨NNT : 2021UCFAC085⟩. ⟨tel-03663791⟩



Record views


Files downloads