A Proposal for Applying Belief, Desire, and Intent Agents toward Automotive Vehicle Energy Management



Authors Abstract
The automotive industry is facing a multifaceted problem of supervisory energy management, computational power, and digitalization. In response, this article proposes the use of agents utilizing the belief, desire, and intent (BDI) framework as a means to flexibly create online vehicle management systems (VMSs). Under such proposal, a community of agents form a vehicle configuration. Each agent represents a vehicle subsystem and contains knowledge specific to its respective hardware. With this knowledge and partial observation over its operating environment, each agent uses the BDI framework to deliberate over its actions. An interaction protocol, which implements a distributed constraint satisfaction problem (DCSP) algorithm, is used between the agents to create sensible emergent behavior of the vehicle. This interaction protocol allows independently reasoning components to produce emergent behavior that is flexible, robust, verifiable, and explainable. In addition, an internal structure on top of the BDI framework is specified which allows an agent to conduct long-term and short-term deliberation asynchronously. A simple, parallel hybrid electric vehicle model is used to demonstrate the application of BDI agents. The agents are tested over the vehicle’s operating envelope to show how independent deliberation and the interaction protocol results in expected behavior and undesirable interactions are avoided. By using agents as modular components, features like dynamic vehicle configurations and persistent intentions are achieved. This article lays the foundation for further studies in the field of applying agents to automotive vehicle control.
Meta TagsDetails
De Vos, S., Frank, T., Abanteriba, S., Sardina, S. et al., "A Proposal for Applying Belief, Desire, and Intent Agents toward Automotive Vehicle Energy Management," SAE Technical Paper 08-09-01-0002, 2020, https://doi.org/10.4271/08-09-01-0002.
Additional Details
Jan 27, 2020
Product Code
Content Type
Journal Article