IPAB Workshop-26/11/2020

Jack Geary

Title: Resolving Conflict in Decision-Making for Autonomous Driving.

Abstract: Conflict arises in Interactive Driving scenarios when two cars fail to coordinate their behaviour during an interactive scenario (e.g. a lane merge). In this talk I will discuss how methods based on the Stackelberg game formulation of the Interactive Driving problem are susceptible to conflict and demonstrate that conflict can cause suboptimal and potentially dangerous behaviour. I will present a theoretical framework for analysing the extent to which such methods are impacted by conflict, and demonstrate the results of applying this framework to several existing approaches modelling interaction between agents. I will introduce a novel approach to modelling interaction between players in a Stackelberg game, and show that it is less prone to conflict than previous techniques.

Alexandros Keros

Title: Sketching Homology and Homology for Sketching

Abstract: Topological Data Analysis offers deformation invariant descriptors that capture the "shape of data", the (persistent) homology groups. While computation of such groups can be performed combinatorially via linear algebraic manipulations of "adjacency"-like matrices, the time and space complexity required deems them impractical for large data volumes and high-dimensional problems. Sketching algorithms provide approximate solutions via "compressing" big-data problems, efficiently. We explore the application of sketching algorithms to improve the computational efficiency of homology, and conversely, we attempt to devise homology preserving "data-reduction" algorithms.

Nov 26 2020 13.00 - 14.00

IPAB Workshop-26/11/2020

Jack Geary, Alexandros Keros

Blackboard Collaborate