Table 1 Summary of representative commercial products and popular demonstrations of obstacle avoidance devices

Name	Sensor	Weight(KG)	Power(W)	Practical User Experience
				Responsive	Reliable	Durable	Usable
WeWALK²⁴	ultrasonic sensor	~0.5	~3	✓		✓	✓
Augmented Cane¹³	LiDAR, camera, GPS, IMU	~1.3	~8	✓	✓	✓
Narayani et al.⁴²	ultrasonic sensors, IR camera	~0.5	~3	✓		✓	✓
InnoMake⁴⁶	LiDAR	~1.5	~20	✓
SuperBrain⁴⁷	RGB camera, point cloud camera, thermal camera	N/A	N/A	✓	✓		✓
Orcam MyEye⁴⁸	camera	~0.5	~25				✓
Horus Eye⁴⁹	LiDAR	~1	~30	✓
SmartSpecs⁵⁰	LiDAR, camera	> 2	~23	✓	✓
Unfolding Space⁵¹	camera	~0.8	~5	✓		✓
ALVU³⁰	TOF sensors	~0.6	~4.5	✓		✓	✓
CaBot⁵²	stereo camera, LiDAR	~25.2	~150	✓	✓
Mocanu et al.⁵³	camera	~0.75	~5			✓

i) Sensor type: Range-based (e.g., ultrasonic) and vision-based sensors (e.g., RGB camera) are commonly used in obstacle avoidance devices. Relying on one sensor type can hardly ensure reliability across diverse scenarios. For example, the WeWALK is susceptible to environmental factors such as temperature, while the Orcam MyEye is constrained by low-lighting conditions. ii) Data processing: Obstacle detection is commonly conducted on-device (e.g., SmartSpecs) or through a wired connection to a laptop (e.g., CaBot). On-device obstacle detection incurs significant delay to process massive data, while carrying a bulky laptop compromises usability. iii) Hardware processor: The typical types include MCUs (e.g., ALVU), CPUs (e.g., SmartSpecs), and GPUs (e.g., CaBot). Due to the compact size and low power requirements, MCUs are commonly chosen as the embedded processors. High power demands of CPUs and GPUs making them less durable for battery-powered devices.

Quick links

Search