Chinese autonomous driving startup Zhijia Xincheng has put its HCT Astra one-stage end-to-end driver assistance system into mass production on the Great Wall Motor (GWM) Dargo HEV, according to Gasgoo. The move marks the first time the system has been deployed on a hybrid-electric vehicle, extending urban NOA (city advanced driver assistance) to models with diverse powertrains.
System capabilities and hardware
The HCT Astra system is built around a single Horizon Journey 6M chip, delivering 128 TOPS of computing power. Through model distillation, compression, and co-optimization of algorithms and software-hardware, the system maintains core decision-making capabilities while using a streamlined model, with performance comparable to chips rated at 200 to 500 TOPS. It employs a diffusion generative model and a one-stage end-to-end algorithm architecture, with response times under 200 milliseconds, and is designed to generalize across cities and scenarios.
In urban settings, the system supports urban NOA, handling complex situations such as mixed pedestrian-vehicle traffic, non-motorized vehicle weaving, and irregular intersections. On highways, it provides adaptive cruise control, automatic lane changes, and ramp merging and exiting. For parking, it offers automatic parking, remote parking, and trail-following reverse, along with active safety features to identify sudden risks.
Expanding across powertrains
The GWM Dargo is the second production model to feature HCT Astra, following the iCAR V27, and the first HEV model to use the system. This rollout signifies HCT Astra's expansion from pure electric to multiple powertrain types, covering hybrid and other diverse models, and further validates its cross-powertrain adaptability. Zhijia Xincheng reports that its neueHCT series has secured orders from over 16 automakers, with more than 70 vehicle models designated for the system.
