Soft Tests for Autonomous Vehicles

A graphic example of soft target benefits!

There is enormous pressure on the entire automotive industry to achieve and demonstrate unprecedented levels of validation to gain and maintain public confidence in the safety of autonomous vehicles.

Statistics may point to human fallibility being the cause of almost all road accidents, but the switch to a connected robotic environment must ultimately deliver every nano-second of every day on the promise of a guaranteed near-total safety highway environment. Today’s grudging acceptance by the global public of the inevitability of deaths and injuries on the road will not continue in a driverless environment.

But moving towards that state has not necessitated a totally clean technology slate - many OEMs, universities, automotive consultancies and government-backed programs have been edging towards it for decades.

An example is advanced testing and measurement systems’ specialist AB Dynamics. In the 1990s, it developed “driving robots”, not autonomous devices but systems that were pre-programed path-following controllers that established a technology and expertise basis for future driverless vehicle testing.

By applying specific, already proven robot driver hardware to a conventional existing vehicle, provides the ability to short cut the early development processes and concentrate on what really matters - software and sensors that can (absolutely reliably) deliver total CAVs (Connected Autonomous Vehicles) that can drive themselves while monitoring and reacting to other vehicles and road situations.

All this has generated the need for test methods to validate CAVs not just by computer but by some physical testing. This can be costly and involve impacts, so the rapid growth in proving ground use of “soft target” testing has become increasingly important.

At AB Dynamics, Commercial Manager Jeremy Ash, explained: “Soft targets are now protecting valuable development vehicles from damage in the event of a collision. CAVs can interact with soft targets that look like other vehicles but are actually a photo-realistic vinyl skin stretched over foam panels. These are mounted on a very low platform that can pass underneath the test vehicle as the foam panels separate during an impact.”

The aerodynamically stable soft target vehicles are self-propelled using electric motors powered by on-board batteries and can reach speeds up to 80km/h. Their interaction with other vehicles is synchronized by telemetry using AB Dynamics’ Synchro software package designed to ensure accurate and timely positioning such as offset head-on impacts.

The company works closely with Euro NCAP and NHTSA, both of whom have adopted the Soft Car 360 vehicle target as the Global Vehicle Target for harmonized ADAS testing.

CAVs have to interact with all types of vehicle but also other road users, including pedestrians, cyclists and motorcyclists. In order to enable testing with these more vulnerable groups a new, compact, motorized platform called LaunchPad has been developed by AB Dynamics, which can carry a range of different VRU (vulnerable road user) targets and is robust enough to be driven over by a test vehicle.

"Soft targets are now protecting valuable development vehicles from damage in the event of a collision," says ABD's Jeremy Ash. They are self propelled using electric motors and can reach 80km/h.

Ash said that unlike belt-driven platforms which can only travel in straight lines, LaunchPad can follow complex paths and uses AB Dynamics’ Synchro package to synchronize its motion with that of other vehicles.

AB Dynamics claims to be the only company capable of synching a VRU, soft car target, and real car in an on-track test scenario, enabling all the mobile architecture in a test to be controlled. The company is also exploring the addition of controlled fixed street furniture, such as traffic lights.

As CAV interactions become more complex, testing will necessarily be carried out using real vehicles as well as soft targets. This is leading to renewed interest in robot driver systems, to enable additional real vehicles to be introduced into the test scenarios.

But true CAVs may well have no pedals or steering wheel, so AB Dynamics is looking ahead to prepare for that, stated Ash: “We are concentrating on the software aspects of ‘by-wire’ control using CAN messages sent over the vehicle’s CAN bus. The challenge is gaining access to the CAN protocol which is usually kept very secure by the OEMs for obvious reasons; they do not want third parties taking control of vehicle speed or direction. Our Flex-0 product uses a standard output format which the OEM can readily translate to suit their own CAN protocol.

As simulation methods become more advanced an increasing proportion of CAV testing is likely to take place in the virtual world, but complemented by real world validation. AB Dynamics is covering both approaches, having recently supplemented its lab and track test products with an advanced driving simulator for driver-in-the-loop simulation.

“It’s a logical step to run autonomous software in-the-loop with the sensors of the CAV seeing what appear to be other vehicles and reacting to them,” explained Ash. “By offering systems for virtual, lab and track testing, we can ensure the compatibility of the software and minimize the workload faced by engineers as they move from one environment to the other.”

Anything that reduces the test workload will be welcome, he said, as CAV testing shows no sign of adopting a single global standard.

Through liaison with bodies such as Euro NCAP, NHTSA and China’s CATARC, Ash is aware of both the diversity and the efforts being made to harmonize CAV testing: “Tests such as NCAP, aimed at providing consumer information, have a different focus to government agency tests, which may concentrate on ensuring the relevance or validity of a particular test type.

“Being realistic, you cannot test all the possible scenarios; the skill is in identifying the best sub-set of test scenarios, based on the worst cases, and demonstrating that these are sufficiently wide-ranging to reasonably represent all likely events.”