Autonomous Vehicle Training and Advanced Driver Assistance Teaching Cars by Watching How Humans Drive

This patent describes systems and methods that capture how a human driver looks, listens, grips the wheel, and uses the pedals, then use those recorded human responses to train autonomous vehicles and to improve advanced driver-assistance systems. The applicant and inventor is Ashok Krishnan.

The Problem

Self-driving technology has advanced quickly, helped by faster computers and better sensors such as lidar, radar, infrared, and ultrasound. Yet truly fully autonomous vehicles have not been achieved. One reason is that machines still struggle with the judgement that experienced human drivers apply automatically: where to look, how to react to an unexpected event, and how to weigh an ambiguous situation. Much of that human skill is never written down or measured, so it is hard to transfer into the software that controls a car. There is a need to capture what skilled humans actually do behind the wheel and turn it into data a vehicle can learn from.

What This Invention Does

The system uses sensors trained on the human, not just the road. It captures human eye movement, hearing, hand grip and contact area, and foot positions while a person drives. From these it extracts event signatures, meaning characteristic patterns that correspond to particular human actions and responses, and it correlates them with events and situations detected by the vehicle’s own sensors and by sensors watching the outside environment. These signatures are classified and stored as raw data and as embeddings, a compact numerical form suited to machine learning. The stored signatures are then used to train vehicles to improve their autonomous capabilities or to enhance driver assistance features. The patent says vehicles can be operated by humans and by software in both the real world and virtual or camera-based worlds. It also describes adjusting the field of view during obstructions, combining camera and lidar data, monitoring driver alertness and attention, and scoring and improving driver performance.

Key Features

• Human sensing. Sensors capture eye movement, hearing, hand grip and contact, and foot positions.
• Event signatures. Characteristic patterns of human response are extracted and labelled.
• Sensor correlation. Human signatures are matched to events seen by vehicle and environment sensors.
• Embeddings for learning. Signatures are stored as raw data and embeddings for training models.
• Real and virtual operation. Vehicles can be driven by people or software in real and simulated worlds.
• Driver monitoring. The system tracks alertness and attention and scores driver performance.

Who Is Behind It

The applicant and named inventor is Ashok Krishnan, an individual applicant with an Australian address for service. The application is a divisional filing from an earlier related application.

Why It Matters

Bridging the gap between human judgement and machine control is one of the central challenges in self-driving technology and in the driver-assistance features now common in new cars. Capturing detailed human responses and using them to train models could help vehicles handle real situations more like an experienced driver would, while the driver-monitoring elements address attention and safety today. Seeking protection in Australia positions the work within the local automotive and road-safety landscape.

Related Concepts

• Self-driving car – the technology the training aims to advance.
• Advanced driver-assistance system – the nearer-term feature set the invention improves.
• Lidar – one of the sensors whose data is combined with camera input.
• Machine learning – the approach used to learn from human signatures.
• Driver drowsiness detection – related to the alertness monitoring the system performs.

---

AU 2026201951 was published in the Australian Official Journal of Patents on 2 April 2026 and is open for public inspection. Patent applications represent inventions that are sought to be protected and do not necessarily reflect commercially available products.