In this article, Senior Technology Analyst Noah El Alami compares and contrasts the suitability of depth cameras and lidar modules for use in commercial humanoid robots, analyzing why OEMs differ between choosing a camera-only or hybrid approach.
Humanoid robots have been pitched as a revolutionary technology for accelerating manufacturing, warehousing, logistics, and more. However, to make the transition from conceptual prototypes to commercial machines, humanoid robots require an array of sensors to operate efficiently and safely in their environment, with IDTechEx forecasting that the opportunity for sensors in humanoid robots could reach over US$6.59 billion by 2037 with a CAGR of 21.4%.
To be more specific, humanoid robots rely on perception sensors to provide information about their environment from a distance, allowing the robot to map its surroundings and potential obstacles, and make safely make decisions about its future actions. A point of contention in the current humanoid robot market is whether cameras alone will be sufficient for humanoid robots to operate, or whether lidar (also known as LiDAR, light detection and ranging) will be needed in combination. This article will delve into this question using insights from IDTechEx's Sensors for Humanoid Robots report.

Benchmarking the relative strengths and weaknesses of lidar and depth cameras for humanoid robots. Image source: IDTechEx.
Two Design Philosophies: "Pure Vision" vs Hybrid
In the current humanoid robot market, there is an ongoing discussion between two different design philosophies: the "pure vision" approach of using only cameras, or a hybrid approach where cameras and lidar modules are combined for mapping and navigation. Tesla is a high-profile proponent of the "pure vision" route, arguing that their autopilot camera technology will be enough for the Optimus humanoid robots to safely navigate their environment.
Meanwhile, many prominent developers of humanoid robots opt to include one or more lidar modules in current humanoid robot designs. The main argument for lidar is that its depth sensing and mapping capabilities are less affected by low or changing ambient lighting, which can disrupt depth cameras. Currently, most trials of industrial humanoid robots take place in factories or warehouses, which are indoor environments with consistent artificial lighting. Deployment of humanoid robots into warehouses and factories will remain as two of their most common environments for the next decade according to IDTechEx research, but as robots move into applications such as security or home-use they must be adaptable to less predictable outdoor settings or shared human-robot spaces.
Advances in 3D Sensing with Depth Cameras
Bringing the focus back onto camera technologies, a key area of innovation for sensors for humanoid robots is in the advancement of 3D or depth cameras. The main approaches to acquiring depth information with cameras are active or passive stereo cameras, structured light, or time-of-flight (ToF) sensing. Particularly relevant to humanoid robots are the two types of stereo cameras.
In passive stereo cameras, the images taken by two image sensors spaced some distance apart is processed to map a 3D image of the environment and nearby objects. Meanwhile, in active stereo cameras, a projecting element, typically an infrared laser, scans across the field of view to assist the two stereo cameras in calculating the distance based on their different views of the landscape. In either case, recent advances in machine vision and on-board AI models have allowed camera-based perception models to map and identify objects and environments with increasing accuracy.

Overview of the types of sensors deployed in humanoid robots. Image source: IDTechEx.
Automotive Influence and Market Outlook
Finally, it is worth noting that there is also significant overlap between the perception sensors used for humanoid robots and automotive, particularly in advanced driver assistance systems (ADAS) and autonomous vehicles. Both systems have similar requirements: to dynamically sense and predict their environment in order to make safe and effective decisions about future actions. Similar technologies and key players can be found supplying both industries, such as Sony, Hesai Technology, Livox, and Ouster, and in the case of a company such as Tesla developing both autonomous vehicles and humanoid robots, there is significant overlap in the development and design of sensors for both applications.
Whatever the design philosophy chosen by humanoid robot manufacturers, it is clear to see that their commercial rollout has been enabled in part by advances in sensing technologies. Depth cameras continue to improve in accuracy and reliability through both hardware improvements and on-board processing and AI. At the same time, lidar modules have also decreased significantly in cost over recent years, making them more suitable for the price point of a commercial humanoid robot. The full report on "Sensors for Humanoid Robots 2027-2037: Technologies, Players, Forecasts" goes into further comparisons and benchmarking of depth cameras and lidars, as well as comparing them to radar and ultrasonic sensors, with 10-year market forecasts for 11 different types of sensing technologies.
For more information on this report, including downloadable sample pages, please visit www.IDTechEx.com/HumanoidSensors, or for the full portfolio of humanoid robot-related research available from IDTechEx, see www.IDTechEx.com/Research/Humanoids.