Historically speaking, driving simulators have always attempted to connect real people with imaginary vehicles. The fundamental principle is not new. But the technology used to make this happen is certainly a moving target.
Over the last thirty years or so, driving simulators have taken on many, many forms – everything from small-scale gaming / entertainment systems, to the large-scale systems used by vehicle manufacturers for their product research and development activities. These days, anyone brave enough to type the words ‘driving simulator’ into an internet search engine will still be bombarded with all sorts of possibilities, simulators with wildly different form factors, features, and price tags. It’s easy to imagine that anyone seeking to learn about driving simulator technology might become confused by all these possibilities.
The wide variation in driving simulator designs has mostly come about due to the wide variety of use cases, and the adoption of emerging technologies. For example, some simulators simply aim to provide display monitor graphics from the drivers' seat perspective to allow visual participation in a simulation, while others aim to engage all the senses with various motion systems, projection graphics, sophisticated audio immersion, and so on.
So, in answering the question posed in the title of this article, one could rightly reply that Driver-in-the-Loop (DIL) simulators are "driven" by external design requirements and available technologies. But, to be honest, I was thinking of this question in more literal, human-centric terms. More specifically: In the proposed new world of driver-less cars, what is the relevance of Driver-in-the-Loop (DIL) simulators at all, of any shape or size?
Man Meets Machine
Driver-in-the-Loop "Driver" = a real person
When a human is required to control the trajectory of a vehicle, we can map it as the following, in control system engineering parlance: The human driver is the ‘control system’ for the ‘plant’ that is the vehicle. The control system (the human driver) provides inputs – throttle, brake, steering, gear selection, tuning selections, etc. – that are injected into the plant (the car), and the plant responds accordingly. The plant provides feedback to the control system (sensory information for the human driver), and the control system responds as needed. And around it goes, with the driver – a real person, in this case – "in the loop."
Driver-in-the-Loop "Driver" = a real person + on-board assistive systems
Enter Advanced Driver Assistance Systems (ADAS). This case, at first glance, appears to violate the traditional human-as-driver case. But, in fact, it does not. This case simply introduces an additional control system for the plant, running in parallel. The control systems (the human driver + physical actuators) provide inputs - throttle, brake, steering, gear selection, tuning selections, etc. – that are injected into the plant (the car), and the plant responds accordingly. The plant provides feedback to the control systems (sensory information for the human driver + sensor measurements and processing logic to physical actuators), and the control systems respond as needed. And around it goes, with the driver – still a real person, in this case – "in the loop."
With ADAS, the primary disruption to the traditional product development / evaluation cycle is due to the purposeful interruption of the human driving (control) tasks, and the potential impact on subjective and objective assessments. Does the lane-keeping implementation inform the human driver in a way that will be acceptable in the marketplace? Is the blind spot detection implementation informative or disruptive to the driving experience? And so on. One might even argue that the value of the human "in-the-loop" component is directly proportional to the level of intervention from the assistive component, i.e. more ADAS warrants more DIL simulation and assessment, rather than less.
Artificial Intelligence (AI)
Driver-in-the-Loop "Driver" = AI, with passive human observer
With driver-less (autonomous) cars, even at the SAE Level 5 conceptual extreme, we still find ourselves in a similar paradigm. In this case an artificial intelligence (AI) control system is in place, and a human occupant is along for the journey. The AI control system provides inputs – throttle, brake, steering, gear selection, tuning selections, etc. – that are injected into the plant (the car), and the plant responds accordingly. The plant provides feedback to the control system (sensor measurements and processing logic to physical actuators), and the control system responds as needed. And around it goes, with the driver – an AI system, in this case – "in the loop."
A valid question is this: At Level 5, with humans completely relieved of traditional driving (control) tasks, is there any need for human participation in vehicle developments – via DIL simulators, or anything else? Well, if the autonomous vehicle in question is simply transporting goods, then perhaps the answer is no. But if the vehicle is a people carrier, then everyone on board will still be the receiving sensory feedback from the vehicle, like it or not. Even if the occupant(s) have zero opportunity to influence the vehicle's trajectory control system, they will still have a subjective experience and opinion. So again, as in the case with ADAS, one might argue that the value of the human "in-the-loop" assessment is directly proportional to the level of intervention from assistive technologies, i.e. more AI may warrant more DIL simulation and assessment, rather than less. There may even be a step increase in the human assessments required to cover Level 2 – Level 4 autonomous developments, where the "driver" definition is further complicated by handover assignments, and testing safety issues force more simulation into DIL labs, etc.
Mapping the Future
All the above – including ADAS and even speculative AI technology – has been at the core developing cars, as consumer products, for the last ~125 years. Think of cruise control. Even though available technology changes rapidly on both the vehicle product side and the product R&D side, the vehicle development script is quite stable. Prototype vehicles and simulation models are built. Expert human evaluators 'drive' them both to assess various performance characteristics (ride, handling, steering, and so on). Objective data is collected alongside subjective assessments, and it's all compared to baseline acceptance criteria and branding/excellence criteria. Changes are verified by simulation and with prototype vehicles, as time allows, before signing off developments into mass-production space.
Driver-in-the-Loop (DIL) simulators have responded accordingly, providing mirrored ‘virtual test driving’ experiences and evaluation opportunities – no matter who's actually "driving" in a trajectory control sense. The advantages of DIL simulator labs are related primarily to efficiency gains, and Return on Investment (ROI) has been associated with getting real people in-the-loop, early and often, with imagined hardware to shrink or eliminate real prototype testing rounds.
The hidden key may be recognizing that human assessment and acceptance will remain at the center of vehicle developments related to human transportation, whether the driver is a person, a person with assistance, or an artificial intelligence. If human evaluators continue to interact with virtual cars and environments via Driver-in-the-Loop (DIL) simulators, we can likely look forward to some fascinating – and enjoyable! – vehicles in the coming years.
Ansible Motion’s DIL simulators are deployed around the world for vehicle development work. To learn more about our unique approaches and technologies, download our FREE white paper, “10 Advantages of Ansible Motion DIL Simulators”.