Salvaging the MaaS Opportunity: Developing Robotaxis for Use in the City

For any disruptive technology, targeted at any market, there must be a “killer app,” a concrete use case with a well-defined ROI to dictate investment, M&A, and go-to-market strategies. For autonomous vehicle technologies, that “killer app” is the Mobility as a Service (MaaS) opportunity. While improving driver safety through better obstacle detection and collision avoidance can be achieved with active safety/ADAS technology, only a comprehensive autonomous vehicle system can kickstart MaaS into the mainstream, driving down the cost per mile of shared mobility modes below that of conventional car ownership.

COVID-19 has proved a potent accelerant to the process of rationalization and consolidation in the AV ecosystem, with severe consequences for a long tail of AV software developers built on a premise of “robotaxi by 2020 or bust.” This begs the question: is there any future for MaaS and its foundational robotaxi technology, or should AV technology developers return their focus to semi-autonomous functions targeted at the traditional passenger car business?

Not according to Intel’s Mobileye, which has valued the MaaS opportunity at US$160 billion by 2030, more than 3X its estimation of the value of the self-driving system market by the same year, an expectation matched by Intel’s US$900 million acquisition of Moovit. In Intel’s own words (in late 2019), “MaaS will govern the self-driving productization pace. Consumer AV Market will be timed by self-driving system productization and consequent cost/value optimization steps within MaaS. Developing MaaS and driving it to quick convergence is critical to secure our self driving system product fit, and to dominate the consumer [i.e., passenger vehicle OEM] AV ramp up ahead of the industry learning curve.”

There are two key takeaways:

• MaaS Needs Autonomous Vehicle Technology: As demonstrated in the ABI Research report Smart Mobility Maintenance: Modular Hardware, OTA Updates, and Prognostics (AN-2601), fully autonomous driving will be the foundation of low-cost, and therefore widespread, adoption of MaaS.
• Autonomous Driving Needs MaaS: Only by scaling and developing AV technology in the context of robotaxis/MaaS can these technologies then be deployed on passenger vehicles at a viable price point.

Given the importance of MaaS to the success of AV technology, it is vital to remember the context in which MaaS will flourish: the city. Therefore, AV developers must continue to target the city as the operating environment for their systems. At a technical level, this means configuring their systems with the necessary compute and sensing technologies to ensure robust 360-degree perception with low latency and high resilience to typically complex urban scenarios. Beyond the technical requirements of a system that can safely navigate the complex urban environment, there are a number of additional steps that AV developers can take to maximize their chances of a successful city deployment:

• Engage with IEEE P2846: One of the biggest barriers to the rollout of autonomous vehicles in cities is risk because driving is a multi-agent problem, involving a variety of decisions and counter-decisions, maneuvers and counter-maneuvers by all of the road users involved. In the city context, these road users include not only other vehicles but also vulnerable pedestrians and cyclists, moving in multiple directions in scenarios involving various occlusions, etc. It is important that AV developers and city authorities move beyond naïve sentiments of autonomous vehicles ending all accidents, and work together to define a framework that will strike the right balance between risk and efficiency in robotaxi deployments. IEEE P2846 is based on Mobileye’s RSS model, a deterministic framework that allows for regulators and legislators to have a role in defining the variables that will inform the balance between risk and efficiency, without involving regulators and legislators too deeply in defining the hardware and software specifications of autonomous vehicles.
• Focus on Validation: Demonstrable and testable safety of AV systems is key to moving beyond the current rut of prototyping and trials into commercial deployments. Long gone are the days when a well-funded startup could present a black box AV software stack with the promise that it would surely be safe!
• Digital Twins/Simulation: While deterministic safety models are the best way to provide verifiable evidence of the safety of an autonomous system, simulation tools have also proved useful in the development and testing of autonomous vehicle components, allowing for rapid prototyping in the digital domain. At the same time, many city governments, such as Singapore, have begun investing in digital twins of their cities in order to boost resilience, help identify potential bottlenecks in the event of a wide-scale evacuation, assess the environmental impact of approving a new building to be constructed, etc. Having developed these digital twins, these models can also help autonomous vehicle deployers to prototype and test vehicles in the digital domain on a specific recreation of the actual deployment city. This is particularly relevant in cities with known “problem areas,” road features that are known to cause complications for all road users.

• Understand City Government Objectives for AVs: Every city government or municipal authority has its own objectives to achieve with the introduction of AVs, as well as fears about the potential costs. Objectives typically include improved safety and personal mobility of vulnerable demographics, while fears also include the cannibalization of public transit by robotaxi MaaS services and the related potential for increased congestion. Demonstrating how AVs can help meet city government objectives as well as a solid strategy for deploying within the existing framework of public transit modes is key to a successful deployment.

The success of autonomous vehicle technology, therefore, relies on effective collaboration between technology developers and city governments. Nevertheless, it is important for AV developers to bear in mind that the onus remains on them to facilitate the deployment of their technology. An important lesson can be derived from the difficult history of the V2X market. With V2V applications being so dependent on a double coincidence of two V2X-equipped vehicles in proximity, V2X adoption has been hampered by low first-mover advantage. The hope for the V2X market was therefore pinned on the deployment of V2X-connected infrastructure by national and city governments, enabling a host of V2I applications to improve first-mover advantage.

Suffice it to say, this market dynamic did not exactly manifest as hoped. Why should cities with stretched budgets commit significant resources to enable an unproven technology for the sake of the automotive industry? If a safety focus on reducing collisions did not save V2I, it can’t be expected that city governments will come to the rescue of AVs on their own initiative and at the taxpayer’s expense.

Perhaps the biggest inhibitor of V2I deployment has been the presence of two competing and incompatible wireless communication protocols, with governments unwilling to risk backing the wrong technology. Similarly, AV developers must engage with governments in a way that does not require them to rule on technology aspects, but allows them to maintain a certain level of control over the pace and scale of technology deployment.