XPeng has unveiled a significant upgrade to its intelligent vehicle system with the second-generation VLA (Vision-Language-Audio) model, introducing the revolutionary ‘Master Agent’. This new iteration aims to transform the in-car experience by allowing the vehicle to not just understand commands, but to truly comprehend user intent and orchestrate complex actions.
From Understanding to Action: The Master Agent
At a recent AI sharing event, XPeng detailed how the enhanced VLA model goes beyond just improving intelligent assisted driving. The company has opted to re-engineer the vehicle’s core computing system using a robotic approach, culminating in the introduction of the Master Agent. This innovation merges the VLA system with the VLM (Vision-Language-Model) cabin, pushing the vehicle’s capabilities from merely ‘understanding a sentence’ to ‘comprehending what the user actually wants to accomplish’.
Under the Hood: Omni Multimodal Model
The Master Agent is built upon XPeng’s proprietary Omni multimodal model. This advanced foundation allows the system to interpret natural language and ambiguous semantic meanings. Crucially, it can automatically break down user intentions into actionable tasks. The Master Agent then coordinates various specialized ‘agents’ responsible for intelligent driving, chassis control, cabin experience, and vehicle body control. This intricate coordination creates a closed loop, transitioning seamlessly from understanding to execution.
Real-World Demonstrations
During the event, XPeng demonstrated the Master Agent’s impressive capabilities. One notable example showcased the ability to control the vehicle to pull over to the side of the road simply by voice command. Furthermore, the second-generation VLA model demonstrated its prowess in understanding vague requests. For instance, a user could describe a desired store using ambiguous terms, and the system would locate the store, plan the route, and initiate navigation.
This advancement signifies a major leap in human-vehicle interaction, moving towards a more intuitive and proactive driving experience where the car acts as a true intelligent assistant.
