Transportation

Ottopia To Provide Remote Assistance For Motional Robotaxis


Contrary to the many public statements and tweets of the CEO of a certain electric vehicle manufacturer, level 5 automated driving systems aren’t likely to become a reality any time in the foreseeable future. In fact, increasingly, many ADS developers are skeptical that it will ever happen. For those not familiar with L5, it refers to an automated vehicle that can operate anywhere and in any conditions. That means automated vehicles (AVs) are going to need help from humans from time to time. That’s why Motional is teaming up with Israeli startup Ottopia. 

As companies begin to prepare for more deployments of AVs without safety operators over the next few years, one of the many pieces of the complex puzzle they have to put in place is some sort of remote assistance or teleoperation capability. This is expected to become a requirement as regulators around the world begin to implement rules for AVs, but even before that, the companies creating these systems recognize that it’s something they have to do. Several companies have popped up over the last few years focused on developing such systems including Ottopia. 

Ottopia has developed a system it calls remote vehicle assistance (RVA) to enable operators to provide hints or suggested paths to AVs in the field. Today, most of the test AVs on the road still have safety operators that can intervene when the vehicle doesn’t know how to handle a situation. There are some AVs that are completely driverless including part of Waymo’s fleet in Arizona and some of the Chrysler Pacificas that Motional is testing in Las Vegas. 

Several years ago at CES, I got a demonstration from a Silicon Valley startup called Phantom Auto of their remote driving solution. In that case, an operator sitting in Mountview, California was watching a live video stream from the cameras on the car we were in on the Las Vegas strip and driving the car using a steering wheel and pedals. The video stream was sent through 4 bonded LTE connections. While it worked, it’s not a solution that is likely to scale well when there are hundreds or thousands of AVs roaming around a city. Even with multiple bonded wireless connections, it’s also not a reliable solution. Cellular dead spots are not uncommon and that’s a safety issue. 

At CES 2017, then-CEO of Nissan Carlos Ghosn talked about a collaboration between the automaker and engineers from the NASA Ames research center. This is a problem that NASA has long had to contend with for the rovers they have on Mars. The 20 minute time delay for signals to reach Mars means they can’t remotely drive in real time. That solution relied on a remote operator looking at the sensor signals beamed from the AV and then drawing in a path for it to take around an obstacle such as a construction zone. 

Ottopia’s founding team has a background in cybersecurity, video compression and telecommunications. The company has combined those skills for its platform. 

“If you know what teleoperation is, then you know it’s all about the performance, the delay the quality of the video, and the consistency of the video, can you operate in areas with very bad cellular coverage?” says Amit Rosenzweig, co-founder and CEO of Ottopia. 

Ottopia’s RVA bonds signals across multiple wireless carriers and uses machine learning to analyze all of the signals in real time to determine which are likely to be the best for the next few seconds. The system was benchmarked by BMW against three other providers over a six month period and it provided the best overall results including low latency and high bandwidth even under challenging network conditions. This setup allows Ottopia to allocate the video streams dynamically across multiple data modems so that the remote operators can see what is around the vehicle and the commands they send back get through to the vehicle. 

While Ottopia can enable full remote driving capabilities, this is not the preferred solution. For safety reasons, remote driving is only being pursued for very low speed applications. Instead the focus now is on something more akin to what NASA does with providing remote guidance and hints to the vehicle. For example if the AV is stuck behind some obstruction like a double parked or disabled vehicle, the remote operator can tell it that it is ok to cross a double yellow line to pass. 

Ottopia has already integrated its system with five different ADS software stacks including Motional and it will be providing the remote assistance capabilities when Motional launches its multi-city service with Lyft

LYFT
beginning in 2023. Motional has already been providing ride-hailing services in Las Vegas through Lyft since mid-2018 as well as services in Singapore since 2016. Motional will begin using the Ottopia RVA in those cities in the fourth quarter of 2021.

Ottopia will be establishing control centers in the areas where the AV fleets are operating in order to help minimize latency of communications and improve reliability. Initially, the ratio of operators to vehicles will be kept relatively low until the reliability of the AVs improves and the load on operators is reduced. Over the next several years, Rosenzweig expects the ratio to grow to about 10 AVs for each operator and in the long run grow even higher. 

Part of the reason for keeping the ratio relatively low is that situations in the real world can change quickly. The remote operators will have to step in for each AV that encounters the same obstruction rather than copying the same instructions to the whole fleet. 

Rosenzweig explains, “It sounds very appealing to do so but it’s a bit risky because even if a few seconds went by between vehicle number one and number two, you’re not completely sure that the solution should be the same. Even though. Supposedly it’s a construction zone, but maybe the actual problem was a double parked car in the construction zone, and that’s the problem and therefore it has a specific solution.”

So even as robotaxis and automated delivery vehicles are deployed in the coming years and drivers are replaced by software, there is still a human role in this ecosystem. People may be flawed, but we can still often make better decisions than machines.



READ NEWS SOURCE

This website uses cookies. By continuing to use this site, you accept our use of cookies.