The top sensors used in autonomous delivery robots

Autonomous delivery robots have been a long-time fantasy, but they are now in real-world development and production. Sensors are the imperative technology to their functionality.

This article looks at the most common sensor types in robotics used for autonomous delivery and explains how they contribute to the autonomous vehicle systems of tomorrow.

LIDAR Sensors

Light detection and ranging sensors (better known as LIDAR) are at the forefront of autonomous vehicle technology. LIDAR robots use near-infrared light to detect the distance between a transmitter and a reflective ‘point,’ similar to a laser-range finder you can find at the hardware store. However, LIDAR systems create millions of light points per second in a 360-degree field of view, allowing each point to contribute to a ‘point cloud.’

Nuro, a leading autonomous delivery vehicle company, utilizes a LIDAR system on their delivery robots. In conjunction with other sensors, LIDAR allows Nuro vehicles to automatically navigate unknown neighborhoods with ease. The LIDAR sensors can detect pedestrians up to 250 meters away and cars at up to 500 meters away.

These point clouds create a fully measurable, 3-dimensional map of the environment around the LIDAR system and the AV. Using AI neural networks and self-driving algorithms to process LIDAR’s point cloud data, autonomous vehicles are able to create virtual, dimensioned worlds in which they can identify known objects such as cars, people and bicycles, and then navigate through them.

There are drawbacks to LIDAR systems, including sensor shadows in which the transmitted light is not able to see objects behind objects. They also require a high computational demand and are not very energy efficient. Still, they are generally used in many autonomous delivery vehicles in development around the world.

Radar Sensors

Similar to LIDAR, radar uses a transmit and receive methodology to find objects. RADAR, which takes its name from an acronym for Radio Detection and Ranging, transmits radio waves outward in a specified direction. It receives the rebound of those waves to understand the distance from objects.

Radar utilizes much larger wavelengths than LIDAR. In turn, it can detect objects further away but at a much lower resolution. It requires far less compute power than LIDAR, which makes it advantageous for specific functions within an autonomous vehicle. For example, radar is excellent at determining the speed of a vehicle relative to other objects, as well as detecting large objects such as houses, cars and other obstructions.

Generally speaking, radar is used supplementally to complete functions that would be unnecessary to complete with LIDAR sensors.

Thermal Cameras

Thus far, autonomous vehicles have a hard time detecting humans in many different environments. Whether in a bustling city center or a quaint suburban neighborhood, humans are surprisingly good at blending into their environment in the eyes of a LIDAR or radar system. As such, thermal imagers and cameras are used to supplement LIDAR and radar models by providing an entirely different set of signature data.

Given that humans themselves emit light, thermal cameras can be used to help detect if a human or other animal is present but is being misrepresented by call-and-response-based technologies. For example, if a human is sitting on a bench next to a trashcan and that trashcan is in the line of sight of an autonomous delivery robot’s LIDAR, the LIDAR system may only identify the trashcan as it cannot ‘see’ the entirety of the human. Thermal cameras, however, would be able to see the human’s body parts and/or heat signatures that are within view of the camera, thereby helping the AV understand that a human is by the trashcan.

Thermal cameras are also able to detect potential systems threats, such as fires, open manholes, or even freshly poured concrete.

GPS in Autonomous Delivery Systems

So far, the sensors used in robotics discussed here have been very focused on micro navigation such as obstacle avoidance or route designation. However, knowing where an autonomous delivery robot is and where it is going is arguably one of its most fundamental capabilities. GPS (or other global navigation systems such as Galileo or BeiDou) sensors are among the fundamental sensors for helping an autonomous delivery robot get from ‘Point A’ to ‘Point B’.

GPS navigation systems have been around for several decades and are certainly not limited to use within autonomous delivery robots, yet their importance for the industry is very significant. One industry leader, Starship, touts that (in conjunction with other onboard sensors) their GPS sensors are able to help autonomous delivery robots understand their location down to the inch.

Cameras

Of all of the sensor technologies listed in this article, cameras are the oldest and, subsequently, the most advanced. Some autonomous vehicle makers such as Tesla prefer to use cameras as their main source of data (as opposed to LIDAR/radar systems) given their natural human-level understandability. Since humans do not see in LIDAR point clouds, identifying and labeling LIDAR data can make neural network training even harder. Cameras, however, process natural reflected light. Humans are spectacularly well equipped to identify, label and understand camera sensor data, making AI training and data understanding very easy.

Are Autonomous Delivery Robots the Future?

There is a multitude of sensors used by autonomous delivery robots and similar autonomous vehicles. While the sensors in this list are the most common found in popular autonomous delivery robots, there are many other powerful robot navigation sensor types also in use, including ultrasonic, current, voltage, accelerometer, barometer, vibration, temperature, humidity, and hall-effect sensors, to name a few. The progress of this technology is happening at blistering speeds. We all may soon have a fully automated robotic mailman delivering our packages to our door.