Find out the advantages and challenges of autonomous vehicle delivery by jumping to 03.54
Akshat is the CTO & Co-founder at Udelv.
Udelv is an autonomous delivery company co-founded by Akshat. Prior to Udelv, he worked for Apple and Tesla’s autonomous driving teams. Akshat worked on both the hardware, software, and engineering management. Before this, he had worked on several projects at General Motors and caterpillar. In these roles, he started off working on power systems, automated motor drives, and motor controls.
2 Advantages & disadvantages of autonomous delivery vehicles
Autonomous driving presents several advantages and disadvantages in the world of delivery. Akshat talked us through the pros and cons of his podcast.
There are no people in an autonomous delivery vehicle. It’s only carrying cargo. When designers are building a vehicle they need to think about safety, comfort and then power-drain-efficiency in that order (This is called the vehicle dynamics theory). They cannot prioritise comfort over safety for example. But with an autonomous vehicle, they can remove the comfort consideration when building the car. This gives designers more time to build in features that consider other road users and the environment.
70-80% of accidents happen on American roads when drivers are turning left. In many states in the US, you are allowed to turn right at a traffic light even when the lights are red because your vehicle won’t be crossing anyone’s path. Combine these two facts are you create a strong argument for the use of autonomous delivery vehicles on the road. By programming them to only turn right at crossroads allows them to reach their destinations quicker and reduce accidents. Whilst this would get pretty frustrating for a human driver, programming a vehicle to do it would increase safety on the roads and the lack of stopping at traffic lights would see the delivery vehicle get to its destination faster. Spread this tactic across your fleet of vehicles and you save a lot of time and create fewer accidents.
If you examine what happens at the start and end of a typical autonomous delivery the first and last 1000 feet can be very challenging. We see a lot of unstructured and uncontrolled environments e.g car parks, loading bay areas, off-road driveways. The autonomous vehicle being created need to have a semantic understanding of such environments. They also need to be able to dynamically plan for these types of environments and predict the intent of other vehicles and people in the area.
To operate an autonomous vehicle today you need to upload regular map updates to the machine to make sure it still understands the routes around town. This isn’t very sustainable as many of the uncontrolled environments the vehicles need to drive through aren’t covered in many maps. This requires the vehicles to understand environmental cues in real time to be truly effective.
01.01 Routes into a career working with autonomous vehicles.
03.54 How is autonomous deliveries different from autonomous driving
10.55 How restrictive are our current roads for autonomous vehicles?
14.01 When will we see autonomous delivery in Manchester?
17.07 Cutting the red tape around autonomous vehicles.
19.18 Narrow AI vs general AI
26.07 Machines minds, human minds
32.02 Creating tangible goals for autonomous vehicles
Location: Northern Powerhouse