Current ACC systems struggle to achieve anticipative driving
Current adaptive cruise control (ACC) systems, used in conventional and self-driving vehicles, have shown little ability to perform “anticipative driving”, which human drivers employ regularly when they see and react to multiple downstream vehicles. Anticipative driving benefits safety by detecting downstream incidents earlier and reducing the response time required, while also reducing traffic oscillations behind the lead vehicle (resulting in less traffic congestion).
Two significant issues with current ACC systems are that 1) they tend to focus primarily on the car directly ahead of the ACC system, and 2) they rely on prediction of the leader’s future movements via machine learning approaches. While manufacturers are adding sensors to detect vehicles and objects beyond the immediate vicinity of the ACC system, machine learning prediction is still an ongoing challenge and requires extra processing power for implementation. The result is that the limited predictive capabilities of these systems can often make traffic worse by amplifying traffic oscillations and not foreseeing potential issues.
History trajectories outperform machine-learning predictions and require less processing power
This technology’s simple approach to improving ACC’s impact on traffic congestion supplants machine-learning-based prediction by focusing on history trajectories. The invention achieves congestion-friendly driving by recording the history trajectory of the leader and then shifting it into the future horizon to obtain a reference trajectory for the subject vehicle (the “ego” vehicle) known as the Newell trajectory. This Newell trajectory is used as a reference to plan a future ego trajectory that does not amplify the speed oscillations from the leader.
The invention further achieves anticipative driving by simply incorporating more history trajectories from one to multiple downstream vehicles, as shown in Figure 1. This additional design allows ACC to have a longer reference trajectory for planning, which in turn helps ACC to react to more downstream vehicles.
History trajectories have been implemented using a 2019 demonstration vehicle, and the additional design for anticipative driving has also been implemented in simulations using the commercially available code base, which suggests that this invention is compatible with current commercial car models with ACC and would continue to enhance performance as more sensors are added to future ACC systems.
Benefits/Advantages
- Ease of Implementation: Leverages existing sensing technologies to reduce congestion through ACC software changes
- Reduced Processing: Using history trajectories to achieve anticipative driving and reduce congestion requires significantly less processing power than machine learning
- Future Friendly: As automated driving systems advance by adding additional sensors, this data can be added via software to produce more history trajectory data, which can further improve driver anticipation capabilities
Potential Commercial Applications
- Vehicles with adaptive cruise control
- Self-driving vehicles
