Automated Driving System (ADS) Development Process

Development interface

In addition to the functionality, attention must also be paid to the development process of an automated driving system (ADS). The following is a brief overview of the debugging functions that are used in the initial development of an electronic control unit (ECU) to get it up and running.

The event logging section outlines retrieving data from the road to serve development, feature improvements, and other uses of recorded event data. Since the industry has already agreed on these aspects in the Auto Safety Alliance, their work is presented here.

Debugging Features

Some non-intrusive measurements can be made using hardware, while others require software and the overhead that comes with it. The system must support “auxiliary” (non-ADAS) uses of the data.

Conceptually (if not physically) there should be a number of consistent and non-interfering data probes (or ports): the original use of the application, the data logger, the debug interface, and the infotainment system.

Sensor data and application state are inspected differently. Sensor data originates from outside the processor and can be copied transparently. Application data may only be visible through intrusive means. Recording the instruction stream is a separate mechanism from recording the data stream.

The following points should be considered for the debugging interface of the functional module:

• Probe Point

• Every I/O and all application state variables

• Bandwidth

• Equivalent to sensors and applications

The debugging functions that may be discussed are as follows:

• Error injection

• Exception capture

• Internal and external interrupts

• Breakpoints

• Monitoring Points

• Step-by-step execution (high and low level)

• Read and write memory

• Record the entire data plane at full data rate

• JTAG

• Instruction tracing

• Timestamp

• HIL/SIL simulation

• Remote debug port access

• Workloads of compute elements such as CPU/GPU and bus interconnects

• Calculate element power consumption

Event Log

Recording data during an incident is important for many different reasons such as crash investigations, system performance studies, failure analysis, continuous learning, etc. Incident analysis and investigations will help identify lessons learned to enable overall improvements in the industry regarding autonomous driving and driver assistance systems.