Micro autonomy’s many relatives
Rumblings about the combination of micromobility + self-driving tech have gotten louder in the last several weeks. One can examine this from two angles:
- Research the vehicles that already exist (read this post)
- Analyze how a micro-autonomous vehicle would function across the self-driving tech stack (stay tuned)
First, a few overall thoughts…
- Commercialization is unsolved. Shared fleet rebalancing is the likely answer, but more on that elsewhere.
- Aesthetics vary widely (but matter). Good: Lit Motors. Bad: AB Dynamics’ Motorcycle. Ugly: Scootbee.
- Terminology is hard. I tried to use consistent, accurate language but found lots of confusion in the coverage. If you want to go deeper, check out Alex Roy’s article, which also explains the terms used in this chart’s x-axis.
Lit Motors C-1
Self-balancing motorcycle-car hybrid
Sensors: Inertia, infrared, temperature and heat sensors, primarily for driving and balancing (source)
Key specs: max speed: 120 mph, range: 200 miles, weight: 800 lb / 360 kg
Deployments: Testing only
Commentary: Lit arrived with a splash in 2010. They have yet to ship their stunning-looking vehicle, nor been acquired (Apple was a rumored suitor). Besides its beauty, the C-1 is noteworthy for its powerful self-balancing abilities: “electronically controlled gyroscopes located under the floor (putting out over 1,300 lb/ft of torque), allow the vehicle to balance at a stop and stay upright in the event of a collision”.
Beijing Lingyun Intelligent Technology 1703
Self-propelled motorcycle-car hybrid
Sensors: unknown
Key specs: max speed: 60 mph / 100 kph, weight: 485 lb / 220 kg, range: 60 miles / 100 km
Deployments: Testing only
Commentary: They say imitation is the sincerest form of flattery, so hopefully Lit Motors feels good about the Lingyun Intelligent Technology prototype “gyrocar”. Vehicle control is interesting- initially, it will sell with a steering wheel, but the prototype includes either a computer mouse or an autonomous mode. The autonomous capabilities are unclear from the promo video but don’t look very robust. In a step forward for EVs, they are promising a swappable 3 kWh lithium battery. Beijing Lingyun does face an uncertain regulatory future in China, where there isn’t a category for this kind of vehicle. Perhaps because of this, they hope to leverage the core technology to produce larger form factor vehicles as well. Learn more
BMW R 1200 GS (modified)
Self-propelled motorcycle
Sensors: unknown
Key specs: weight: more than 500 lb / 227 kg
Deployments: Demo only
Commentary: Intended to showcase emerging technologies, e.g. motorcycle ADAS like steering assistance in corners or automatic emergency braking (AEB). Notably, this bike uses no gyroscopes; only standard brake, throttle and steering controls. It doesn’t appear to feature any notable sensors for perception. Learn more
AB Dynamics/AutoRD BMW C1 (modified)
Self-propelled motorcycle
Sensors: unknown
Key specs: weight: more than 185 kilograms (408 lb)
Deployments: Demo only
Commentary: Calling this bike driverless is a stretch. It can “synchronize its motion with other moving objects”, perform maneuvers like passing that require a combo of steering, braking and accelerating on its own. However, it lacks any sensors for object detection, meaning no prediction (or very good planning) is yet possible. AB Dynamics envisions this eventually being used by commuters, much as a self-driving car. Learn more
Starship Technologies
Self-driving sidewalk robot
Sensors: 9 cameras (3 front, 4 sides, 2 rear), ultrasonic obstacle detectors (front) and radar, 2 IMU, GPS (source)
Key specs: max speed: 6 kph, 3 kph effective, range: 6 km (~2 hrs of operations), weight: 50 lb / 23 kg
Deployments: US, UK, Germany (tested in >100 cities)
Commentary: the best funded of the sidewalk robot companies ($42M), Starship is illustrative of the category. It’s self-driving, but sticks to the sidewalks and has teleoperation capability for any really challenging situations. They’re focused on food delivery and other short distance errands so far. Learn more
Marble
Self-driving sidewalk robot
Sensors: cameras, LIDAR, and ultrasonic sensors (source)
Key specs: speed: “walking speed”, weight: more than 80 lb / 26 kg (source)
Deployments: US- San Francisco (completed), Texas (planned)
Commentary: Marble is the F-150 of sidewalk robots (h/t Reilly Brennan). It’s waist-high, sturdy and has lidar. This does mean its operators will first have to map its entire operating area. Vehicle and operating costs are likely very high; perhaps in part why they’ve apparently struggled to get this deployed. As an aside, the CEO and co-founder Kevin Peterson is a serious autonomous vehicle industry vet- he was a member of Carnegie Mellon’s DARPA team. Learn more
Scootbee
Self-driving scooter
Sensors: Camera (1 front, 1 rear?), Ultrasonic (at least 4 rear?), IMU, DGPS, Compass (source)
Key specs: max speed: 9 mph /15kph (w/ passenger), slower than human walking (self-driving)
Deployments: six month trial in One-North district, Singapore
Commentary: This appears to be the first commercially deployed micro autonomous vehicle, but it’s underwhelming. For one, it’s slow for riders (<10 mph). It’s also pretty ugly. Aesthetics matter, both for gaining new riders and for minimizing vandalism (I recommend Horace Dediu and Oliver Bruce’s podcast on this). It appears to only travel along sidewalks and other pedestrian zones with clear markings (i.e. not roads).
Tsinghua University “self-riding” bicycle
Auto-balancing, self-propelled bicycle
Sensors: 2015 version- SICK LMS111 Lidar, laser-based speedometer (plus a gyroscope); 2018 version- camera, lidar(?)
Key specs: unknown
Deployments: Demo only
Commentary: This is starting to approach a self-driving bicycle. It can steer, brake, accelerate, and balance without a human on board. While it doesn’t have localization or route planning capabilities, the 2018 bike can track and follow a person using cameras and computer vision. The older 2015 version appears to have been remote-controlled only.
The team cleaned up the hardware significantly in the three years between released versions. What’s not clear is whether it can balance when entirely stationary (I suspect not). Learn more about the 2015 and 2018 bikes.
Thanks to Oliver Bruce for his feedback on this post.