Risk thresholds for autonomous collision avoidance

Abstract

An avoidance manoeuvre is decided on a collision probability threshold. The threshold is a compromise: lowering it reduces residual risk but multiplies manoeuvres, and every manoeuvre costs propellant, interrupts observations and itself introduces orbit uncertainty.

The thesis studies how the choice of threshold affects manoeuvre count and propellant use, on a set of virtual conjunctions built to be representative.

Objectives

1. Build a set of virtual conjunctions with realistic distributions of miss distance and covariance.
2. Compute the collision probability for each.
3. Simulate the manoeuvre policy as the threshold varies.
4. Derive the curve between residual risk and propellant spent.

Methodology

1. Generation of the conjunctions by sampling, with parameters taken from public statistics.
2. Probability computation with the classical two dimensional method, and sensitivity analysis on covariance.
3. Sizing of the minimum manoeuvre that brings the probability below threshold.
4. Simulation over one mission year, counting manoeuvres and cumulative propellant.
5. Study of the effect of an underestimated covariance, which is the realistic case.

Expected contributions

1. The residual risk against propellant curve, for a typical mission.
2. A reasoned indication of the threshold, instead of a value taken by convention.
3. Quantification of how much an optimistic covariance shifts that choice.

Tools

Python, public conjunction data and catalogue statistics.