HiSilicon APA and AVP Chipset Solution

Challenges

Automatic Parking Assist (APA) utilizes ultrasonic radars, visual sensors, and ADAS processors to locate appropriate parking spaces. It identifies spaces that are the right size for the vehicle, plans the parking route, and controls the steering wheel, gearbox, and accelerator pedal to assist with parking. APA is more intelligent than traditional parking assist systems, and helps drivers park their vehicles quickly and safely even in narrow spaces. Automated Valet Parking (AVP) enables drivers to park or summon their vehicle with just a tap. Once the vehicle has received an instruction, it either parks itself or drives from its current parking space to the pick-up location. This can be achieved without the need for the driver to operator and monitor their vehicle, since AVP is a more automated system than APA. Despite that the automated parking systems are highly convenient, they remain under-utilized due to technological challenges that still need to be addressed.

Inadequate perceptive capabilities

As vehicles are expected to deal with more complex scenarios, for example, planning safe driving routes that avoid obstructions and pedestrians, the environment perception capability is particularly important.
Need for better algorithms

Vehicles are required to quickly identify appropriate parking spaces among diverse options, and so high-performing algorithms are needed to ensure automatic obstacle avoidance.

Highlights

HiSilicon APA and AVP solutions support 8-channel ISP capabilities, and use a short-distance camera and an ultrasonic sensor to display a 360 degree view of the area surrounding the vehicle in real time. With 4 TOPS of AI compute power, the APA and AVP can automatically search for parking spaces and identify hidden obstacles, resulting in safer and more accurate parking.

Multi-channel video processing

A single chipset supports 4 to 8 camera inputs, and the ISP is independently configured, supporting flexible camera configuration with a resolution of 2 to 8 megapixels and covering all parking scenarios.
Robust AI compute power and development platform

Our solution provides 4 TOPS of AI compute power, efficient algorithm tuning tools, and on-device self-learning platforms.
Surround view

Our chipsets are embedded with hardware acceleration modules for image stitching, delivering a three-dimensional 360 degree view of the area surrounding the vehicle in real time.