Ways to Save the Future of Robo-taxis

Artificial-intelligence algorithms, trained through rigorous testing, can analyze a self-driving car’s data and tackle the most complex driving scenarios.

Marie Hattar

December 5, 2023

5 Min Read
Waymo SF_Exterior
Waymo partners with Chinese automaker Geely to put AV technology into Zeekr BEV.

California’s Public Utilities Commission (CPUC) offered a glimpse into the future of driverless cars when it approved two autonomous-vehicle “robo-taxi” companies' applications to expand operations in the Bay Area.

The ruling allowed Google’s Waymo and General Motors’ Cruise to operate self-driving car rides at any hour of the day in San Francisco. Since these projects got the greenlight in August 2023, wider use has been rocky and major setbacks have left many wondering about how robo-taxi technology works and when they will actually be ready to take on the responsibility of safely transporting humans. It has also left some to ponder which tools in these companies’ automotive arsenal can help steer self-driving cars back onto smooth terrain.

Three significant innovation opportunities will create a brighter future for autonomous vehicles.

Restore Public Trust With Rigorous Testing and Simulation

Subjecting autonomous vehicles to a rotating cast of distracted drivers and jaywalking pedestrians is exactly what’s needed before they take on the responsibility of transporting human lives and there’s no better way to do this than through simulation. This all comes back to artificial-intelligence algorithms, trained through rigorous testing, that can analyze a self-driving car’s data and tackle the most complex driving scenarios.

For starters, robo-taxis have a variety of onboard sensors, including radar, lidar and cameras, to provide the car with an uncanny ability to “see” all around them. Combining the latest server-grade central processing units (CPUs) and graphics processing units (GPUs), this onboard computer takes information from onboard sensors, analyzes surrounding objects and identifies a safe route forward. However, only practice makes perfect...many, many miles of practice…and again, that’s where simulation holds its value. 

While there is no substitute for testing on tracks and public roads, simulation helps log many more miles that otherwise would not have been accounted for, and it does so in the fastest way possible. Waymo, for example, claims its cars have driven more than 20 million miles (32.2 million km)on public roads and 20 billion miles (32.2 billion) in simulation. That’s one thousand times as many miles in simulation. Through simulation, autonomous vehicles can practice responding to complex scenarios, including edge cases, which are rare but potentially disastrous scenarios that can lead to accidents. Edge cases would also be much more difficult and dangerous to test in the real world.

During simulations, digital twin technology, or digital representations of real-time data that aid in building predictive models, is often used to test a wide range of automotive features, including radar, cellular vehicle-to-everything (C-V2X), in-vehicle networks (IVN) and cyber-security. Many automakers are integrating multiple digital twins to test vehicles in complex scenarios where they can detect and differentiate multiple objects in dense scenes.

By emulating real-world driving situations in the lab and manipulating the variables in each test, automakers can log many more test miles and drastically boost the likelihood of getting it right while minimizing risk.

Enlist the Help of C-V2X

C-V2X also can help robo-taxis speed toward wider adoption. This connected mobility platform allows vehicles to interact with other vehicles, pedestrians and infrastructure to improve safety. The cellular-based technology enables communication between connected devices, such as a vehicle’s electronic control units (ECUs), roadside infrastructure and even apps running on smartphones. Essentially, C-V2X enables messages conveying different types of information to be sent and received. A few examples include critical alerts to avoid collisions and informing drivers of an emergency vehicle coming from behind or a pedestrian crossing ahead.

C-V2X has two modes of operation: direct mode and network mode. The direct mode allows for direct communications, such as vehicle-to-vehicle (V2V) and vehicle-to-pedestrian (V2P), and it operates without the need for a connection to a cellular mobile network. Network mode utilizes a cellular network to convey messages over a much wider area for applications that typically do not require very low latency. Using radio technology, C-V2X essentially allows cars to “see around corners” to communicate the direction of travel, location and speed of surrounding vehicles, bicycles and pedestrians. Onboard processors use specialized algorithms to analyze this information, predict the movement of a car’s surroundings and plan the best course of action.

In situations like a music festival, where concertgoers on their cell phones are overtaxing cellular networks, C-V2X direct mode could be a promising path forward for robo-taxis. These scenarios also can be tested in simulations where radio frequency (RF) conditions could be manipulated to mimic or even supersede the real world, to determine precisely how the vehicle performs in extreme conditions.

Test Software to Ensure There Are No Delays in Decision Making

Automotive in-vehicle network (IVN) testing is another opportunity for robo-taxis. To ensure optimal design, functionality and safety of connected cars, automakers and their suppliers can use comprehensive IVN test solutions to validate devices, systems and applications, as well as the car’s entire network. 

Marie Hattar.jpg

Marie Hattar_0

By using IVN and faster processors, robo-taxis can keep internal communications as efficient and lightweight as possible. Through rigorous testing of system design, robo-taxis will have the ability to intuitively analyze data coming from all sensors, such as radar, lidar, cameras, Bluetooth, ultrasonics and, perhaps, even C-V2X and can train its algorithms – with the help of artificial intelligence – to make super-fast and accurate decisions on what’s most urgent, such as braking for an oncoming pedestrian or avoiding a collision. Some of that data also can be offloaded to the cloud when appropriate. By using IVN testing to ensure that the sensor fusion, or the process of collectively taking inputs from sensors to interpret environmental conditions, is reliable and effective, self-driving cars will be even better equipped to handle the most difficult scenarios.

Marie Hattar (pictured, above left) is senior vice president and chief marketing officer of Keysight Technologies, a provider of electronic design and test solutions,

Subscribe to a WardsAuto newsletter today!
Get the latest automotive news delivered daily or weekly. With 5 newsletters to choose from, each curated by our Editors, you can decide what matters to you most.

You May Also Like