Navigating Real-World Challenges: A Quadruped Robot Guiding System for Visually Impaired People in Diverse Environments

Title of the Paper

Navigating Real-World Challenges: A Quadruped Robot Guiding System for Visually Impaired People in Diverse Environments

Bibliographic Information

• Research Domain: Assistive Technology, Robotic Navigation Systems
• Keywords: Visual Impairment, Orientation and Mobility, Assistive Technology, Navigation, Robotic Guide Dog

Research Background and Issues

• What problems or challenges did the authors identify?
  Visually impaired individuals (BVI) face significant challenges in independently navigating unfamiliar environments, including wayfinding and obstacle avoidance. Mainstream tools such as white canes and smart assistive devices have limitations, such as the inability to perceive environmental information in advance or causing cognitive overload in complex environments.

• Why is this issue important?
  According to the World Health Organization, there are 290 million people globally with visual impairments, including 43 million who are completely blind. Approximately 90% of visually impaired individuals require assistance to leave their homes, and those who can navigate independently often only manage fixed routes. Developing more efficient assistive navigation tools is crucial for enhancing the independence of visually impaired individuals.

• Research Motivation and Related Work
  Existing electronic assistive devices, such as smart canes and wearable navigation systems, feature advanced sensing technologies but provide ambiguous, untimely feedback that increases cognitive load, leading to low adoption rates. Recently, robotic navigation systems with force feedback have shown potential but are primarily limited to indoor flat environments. This paper proposes a quadruped robotic navigation system, RDog, to address the complex navigation challenges faced by visually impaired individuals.

Solution

• What methods or solutions did the authors propose?
  The authors designed the quadruped robot RDog, which combines advanced mapping and navigation systems to provide force feedback and predictive voice feedback, assisting users in navigating and avoiding obstacles in both indoor and outdoor environments.

• What are the innovative aspects of the solution?

  1. Utilizing a quadruped robot instead of traditional wheeled robots, enabling adaptability to various terrains.
  2. Integrating force feedback and relevant voice feedback to offer visually impaired users a more intuitive and clear navigation experience.
  3. Active obstacle avoidance and path planning functions significantly improve navigation efficiency and safety.

• What are the implementation steps and key technologies used?

  1. Hardware Design: Employing the Unitree Go1 EDU quadruped robot equipped with five sets of fisheye stereo cameras, ultrasonic sensors, and Livox Mid-360 LiDAR. Navigation algorithms are processed using the NVIDIA Orin computing unit.
  2. Map Construction: Generating multi-layer maps based on visual LiDAR, including 3D point clouds, 2D occupancy grids, and behavior node layers.
  3. Navigation Functionality: Utilizing hierarchical path planning algorithms (A*) and dynamic window obstacle avoidance models to ensure smooth and safe navigation.
  4. Interaction Design: Equipped with a rigid handle for force feedback and joystick and button controls, allowing users to adjust speed and pause autonomously. Voice feedback is used to notify users of directional and terrain changes in advance.
  5. Technical Validation: Conducting experiments in real-world environments to compare the robotic system with white canes and smart canes, highlighting differences and advantages.

Research Outcomes

• What specific outcomes were achieved?

  1. RDog significantly improved navigation efficiency, requiring less time to complete tasks compared to white canes and smart canes, with reduced errors and collisions.
  2. Subjective evaluations showed that RDog imposed the lowest cognitive load (NASA-TLX metrics) and provided a positive user experience (high SUS scores).
  3. Navigation performance in complex environments, such as restaurants and courtyards, surpassed traditional devices.

• What advantages does it have compared to existing solutions?

  1. Adaptability to various terrains, including grass, brick roads, and stairs, which is a first-time validation for robotic navigation solutions.
  2. Active obstacle avoidance and predictive feedback functions significantly reduce users' cognitive load.
  3. Multi-layer map modes enable more efficient path planning and precise localization.

• What are the experimental or evaluation results?

  1. Navigation Time: RDog reduced navigation time by approximately 17 seconds in indoor environments and 61 seconds in irregular environments. It also demonstrated advantages in multi-terrain scenarios.
  2. Smoothness: RDog's path was noticeably more continuous compared to white canes and smart canes, requiring no intervention from experiment personnel.
  3. Safety: Participants experienced the lowest collision rates and maintained navigation paths farther from obstacles when using RDog.
  4. Function Validation: Users successfully completed complex tasks, such as retrieving meals in cafeterias and returning trays, using voice commands, significantly enhancing independence.

• Limitations and Future Directions

  1. The system currently requires pre-constructed maps, limiting its applicability. Future work should integrate universal mapping systems to expand usage scenarios.
  2. The quadruped robot has high power consumption, necessitating optimization of battery life.
  3. Specific terrain validations, such as stairs and steep slopes, have yet to be fully conducted.
  4. The sample size of users is limited; further experiments should recruit more guide dog users for comparative analysis.
  5. Social acceptance has not been thoroughly studied; future research should explore privacy, safety, and interaction impacts on visually impaired individuals and the surrounding public.

Conclusion

The RDog system proposed in this paper significantly enhances navigation efficiency for visually impaired individuals in complex environments. By combining force feedback and voice prompts, it reduces cognitive load and improves user independence. This study provides valuable insights for developing multi-terrain assistive navigation devices and lays a foundation for future research. Future efforts are expected to optimize the design, expand application scenarios, and investigate social impacts to transform the technology into practical products.