30hp - Energy-Based Models for Out-of-Distribution Detection in Autonomous Driving

Introduction: Thesis work is an excellent way to get closer to Scania and build relationships for the future. Many of today's employees began their Scania career with their degree project.

**Background **Autonomous vehicles rely on perception models trained on limited distributions of data. In real-world driving, out-of-distribution (OOD) inputs frequently occur, unusual weather, new traffic signs, rare objects, or sensor faults. Detecting such OOD cases is critical for safe operation, since models often fail silently with overconfident predictions. Energy-Based Models (EBMs) provide a principled way to flag unlikely inputs. Unlike softmax probabilities, which can be overconfident, EBMs assign higher “energy” to inputs far from the training distribution. This makes them suitable for real-time OOD detection in camera-based surround-view driving systems.

**Problem Statement **Can EBMs be effectively integrated into autonomous driving to provide reliable, low-latency OOD detection, improving the safety and robustness of autonomous driving systems?

Objectives

• Implement an energy-based scoring method for camera-based perception models (BEV/occupancy).
• Compare EBM-based OOD detection with baselines (softmax confidence, dropout, ensembles).
• Evaluate real-time feasibility on embedded automotive hardware
• Assess impact on safety by correlating OOD detection with planner overrides and interventions.

Education/program/focus:

Indicate education, program or focus: Masters program on computer science with a focus on AI

Number of students: 1 Start date for the thesis work: January 2026 Estimated time required: 6 months

Contact persons and supervisors:

Mohammad Nazari, Ph.D.,  mohammad.nazari@scania.com

Joanna Fonseca, Ph.D., joana.fonseca@scania.com

Application: Your application must include a CV, personal letter and transcript of grades

A background check might be conducted for this position. We are conducting interviews continuously and may close the recruitment earlier than the date specified.