Cetin Mericli is the Co-Founder & CEO of Locomation, where he leads a team of the world’s foremost experts on autonomous vehicles, robotics, and artificial intelligence.
Previously he was a Special Faculty – Commercialization Specialist at the National Robotics Engineering Center, Robotics Institute, Carnegie Mellon University, where he led the Machine Learning and Perception group.
What initially sparked your fascination with robotics and autonomous vehicles?
I developed an interest in computers as a kid and taught myself computer programming. That quickly led to an interest in getting computers to make decisions on their own. An article about an MIT project on a legged robot learning how to walk sparked my interest in robotics in the early ‘90s, and a few years later another robot project I stumbled upon cemented my interest and I decided to dedicate my career to building intelligent autonomous robots. The rest is history.
From 2013 to 2018 you were leading the Perception and Machine Learning Group at Carnegie Mellon University, with a focus on autonomous driving. What were some of the projects that you were working on?
I’ve been one of the luckiest people on Earth to have the opportunity to spend around a decade at Carnegie Mellon with the true pioneers and trailblazers of robotics and Artificial Intelligence. During my time at the National Robotics Engineering Center (NREC) as a special faculty, I’ve led and contributed to several high-profile projects. They have ranged from autonomous and semi-autonomous ground vehicles for the US Military and the world’s first fully autonomous production system, Caterpillar’s autonomous mine hauling trucks, all the way to assistive autonomy for helicopters and autonomous agriculture vehicles. It has been like drinking from the firehose, and I’ve learned an unparalleled amount of know-how and know-how-not-to.
Could you share the genesis story behind Locomation?
The founding team of Locomation was the Autonomy Group at NREC/CMU, and we’ve been working together on dozens of projects for the past decade or so. We have been very opinionated about how, when, and where autonomous vehicles can be deployed, and we wanted to laser focus our over a century of AI and robotics experience on commercializing AVs. We always had a meta hypothesis saying over time people will commute less and as a result of the see-saw relationship there, “things”, or freight, will have to come to people. Of course the pandemic accelerated that trend by 5-10 years, but this was bound to happen and we wanted to be on top of that. Therefore, we decided to focus on freight transportation as a starting point, and we wanted to address the most salient pain points such as labor shortage, operating costs, and environmental sustainability. All arrows were pointing to semi-trucks and middle mile. However, making an 80,000lbs semi-truck fully autonomous so it can drive itself at 70mph on an interstate is also non-trivial and it will require very rigorous and tedious safety testing and validation with several open questions being active research areas. We wanted to make sure we can launch a commercial product in a short period of time, and then iterate over it. Therefore, we decided to introduce the Human Guided Autonomous Convoy concept to deploy fully autonomous semi-trucks with human guidance in a very efficient and profitable way. That is how Locomation was born.
Could you explain the concept behind Locomation with the two trucks and two drivers?
Locomation’s first product is called Autonomous Relay Convoy (ARC) and it consists of two trucks and two drivers. It is designed to operate on long haul routes, typically around 1000 miles per day. The operating concept is devised to provide a frictionless, end-to-end solution. Two trucks start their journey driven manually by two drivers, through the urban/surface streets, and onto the highway. Once both trucks are on the interstate, the ARC system is engaged, and the second truck goes into full autonomous leader following mode. At that point, the driver in the lead truck stays on duty and engaged, but the driver in the second truck signs off and goes to rest. Every couple of hours, at pre-planned and pre-determined points, the trucks swap positions so both drivers can take turns in being in the lead position and resting. At the end of the interstate stretch, the ARC system is disengaged, two drivers go back to manually driving both trucks through the final mile. This way, we can deliver 2x more freight to 2x farther and 2x faster while reducing the operating costs significantly.
How many miles of data have been collected from the trucks operating on the road?
We do not measure progress just by the amount of data collected or miles driven. We are optimizing for the information content in the data. We are meticulous about how to plan our tests and what to test when, where, and how. Having said that, we have driven tens of thousands of miles with our trucks, and collected many thousands of miles of autonomous operation data from multiple regions.
What type of cost/fuel savings can fleet truck operators expect from using Locomation equipped trucks?
It depends on the actual route, but on average, we are talking about ~8% fuel savings just from aerodynamic drag reduction (and a couple more percent points for autonomous driving efficiency), and doubling the freight capacity for the same resources, hence reducing the associated labor cost by around 50%. Our convoys also reduce the amount of CO2 emissions by over 40 metric tons per year per convoy. As you can imagine, it adds up to a massive amount for large fleets.
Locomation plans on eventually achieving full autonomy in four phases. Could you share with us what these phases are?
We believe a properly designed, multi-phased compounding approach is what will get us to full autonomy in the safest, fastest, and broadest manner. Autonomous Relay Convoy is our Minimum Viable Product that will hit the commercial market first. Soon after, we will release Drone Follower, which is a two truck, one driver configuration where a single driver effectively transports twice as much cargo on routes that can be round-tripped in a day. These two convoy systems will be launched in the next 3 years. Then, we will start collecting very diverse real world validation data from our commercial convoys at an unprecedented scale, so we will make accelerated progress towards the last two phases for driverless deployments. First, we will get to deploy solo driverless trucks on interstate segments only (Hub-to-hub), and then eventually we will have safety and reliability validation for our technology to handle dock-to-dock full autonomy, including driving autonomously on the surface streets. Each phase will first come to a select few routes and at small scale, but we will always have our entire supply-chain running with some form of autonomy, and the stakeholders will be able to start benefiting from the efficiency, quality, safety, and environmental gains as quickly as possible. This staged approach will remove the immense commercial pressure on the industry and also on us, so we can focus on doing our job right rather than having to do it fast.
Do you have a rough timeline for when full autonomy will be achieved?
The first meaningful and wide scale deployment of autonomous vehicles will be in autonomous trucking and will be in the form of human-guided autonomous convoys.
The only things we know about fully driverless autonomous vehicles is 1) full autonomy will be achieved eventually, 2) it will not be in very near term, and 3) we don’t know what we don’t know. I think in the next several years, we will start seeing very small-scale deployments and by the end of the decade, we will start seeing wider use cases, still mostly in freight transportation. Ubiquitous deployment of self-driving vehicles for people transportation (such as robotaxis) is going to take a very long time unless there is a completely unforeseen breakthrough.
Is there anything else that you would like to share about Locomation?
We have been very fortunate to have assembled the most experienced team (average AV experience is over 14 years in the engineering team, average trucking experience is over 30 years in the commercialization team) and we like each other so much that every face lights up when we see each other (unfortunately mostly virtually, these days). We have an amazing culture, and we are very selective about who we add to our team. Having said that, we are always looking for exceptional people to come join us, make us better, and give us the opportunity to learn from them. So, for all the people working on self-driving vehicles for a number of years who are frustrated with the sand trapped progress, and who are obsessed with actually putting self driving vehicles on the roads, check out https://locomation.ai/join-us/
We are Locomation, and we will be the first company to deploy autonomous trucks at scale.
Thank you for the great interview, I look forward to seeing these trucks on the road in the near future. Readers who wish to learn more should visit Locomation.
- Attention-Based Deep Learning Networks Could Improve Sonar Systems
- Cerebras CS-1 System Integrated Into Lassen Supercomputer
- Deepfaked Voice Enabled $35 Million Bank Heist in 2020
- Facebook: ‘Nanotargeting’ Users Based Solely on Their Perceived Interests
- IBM Announces AI-Driven Software for Environmental Intelligence