Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Secure Control of Connected and Autonomous Electrified Vehicles Under Adversarial Cyber-Attacks

Shashank Dhananjay Vyas, Satadru Dey

TL;DR

The paper tackles the vulnerability of Connected and Autonomous Electrified Vehicles (CAEV) to adversarial cyber-attacks that disrupt both sensing and communication. It proposes a secure control framework that augments the CACC-based powertrain with an RL-based defender, trained via Proximal Policy Optimization (PPO), and driven by residuals from vehicle and battery state estimators. Through simulation, the defender maintains the desired inter-vehicle distance and prevents collisions under coordinated attacks on acceleration and battery current, with residuals serving as attack signals to the RL agent. The work demonstrates the feasibility of RL-based cyber-physical defense in CAEV and outlines avenues for extending the approach to broader attack vectors and platoon scenarios.

Abstract

Connected and Autonomous Electrified Vehicles (CAEV) is the solution to the future smart mobility having benefits of efficient traffic flow and cleaner environmental impact. Although CAEV has advantages they are still susceptible to adversarial cyber attacks due to their autonomous electric operation and the involved connectivity. To alleviate this issue, we propose a secure control architecture of CAEV. Particularly, we design an additional control input using Reinforcement Learning (RL) to be applied to the vehicle powertrain along with the input commanded by the battery. We present simulation case studies to demonstrate the potential of the proposed approach in keeping the CAEV platoon operating safely without collisions by curbing the effect of adversarial attacks.

Secure Control of Connected and Autonomous Electrified Vehicles Under Adversarial Cyber-Attacks

TL;DR

The paper tackles the vulnerability of Connected and Autonomous Electrified Vehicles (CAEV) to adversarial cyber-attacks that disrupt both sensing and communication. It proposes a secure control framework that augments the CACC-based powertrain with an RL-based defender, trained via Proximal Policy Optimization (PPO), and driven by residuals from vehicle and battery state estimators. Through simulation, the defender maintains the desired inter-vehicle distance and prevents collisions under coordinated attacks on acceleration and battery current, with residuals serving as attack signals to the RL agent. The work demonstrates the feasibility of RL-based cyber-physical defense in CAEV and outlines avenues for extending the approach to broader attack vectors and platoon scenarios.

Abstract

Connected and Autonomous Electrified Vehicles (CAEV) is the solution to the future smart mobility having benefits of efficient traffic flow and cleaner environmental impact. Although CAEV has advantages they are still susceptible to adversarial cyber attacks due to their autonomous electric operation and the involved connectivity. To alleviate this issue, we propose a secure control architecture of CAEV. Particularly, we design an additional control input using Reinforcement Learning (RL) to be applied to the vehicle powertrain along with the input commanded by the battery. We present simulation case studies to demonstrate the potential of the proposed approach in keeping the CAEV platoon operating safely without collisions by curbing the effect of adversarial attacks.

Paper Structure

This paper contains 16 sections, 17 equations, 11 figures, 1 table.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Research Gaps and Main Contribution
  4. Cyber-Physical Security in Connected Autonomous Electric Vehicles (CAEV)
  5. Operational Description of CAEV
  6. Cyber-Attacks in CAEV
  7. Modeling of CAEV
  8. CAEV Model under Cyber-Attacks
  9. CAEV Model under Cyber-Attacks
  10. Secure Control of CAEV
  11. Overview of the CAEV Defense Approach
  12. Details of the Defender Architecture
  13. Results and Discussion
  14. Case Study 1: Agent Performance under Coordinated Attacks
  15. Case Study 2: Agent Performance under Limiting Attack Scenarios
  16. ...and 1 more sections

Figures (11)

  • Figure 1: (A) An example schematic of a battery-powered platoon under Coordinated Adaptive Cruise Control (CACC) strategy. (B) An expanded view of the local autonomous vehicle controller operating as a part of the vehicle platoon.
  • Figure 2: A schematic of the proposed secure control framework for CAEV system using a reinforcement learning based cyber-defender algorithm.
  • Figure 3: Reference velocity trajectory for leader vehicle.
  • Figure 4: Training progress of the PPO agent: rewards vs episode.
  • Figure 5: Adversarial vector profiles: Acceleration attack, and Current attack.
  • ...and 6 more figures