Autonomous Selection and Printing of 3D Models for People Who Are Blind
Timo Götzelmann · 2018 · ACM Transactions on Accessible Computing · doi:10.1145/3241066
Summary
This paper addresses a significant autonomy gap: while 3D printed models are increasingly used in schools for blind students to understand spatial concepts through touch, the process of finding and printing these models has required sighted assistance. People who are blind could not independently search online 3D model repositories (which present results visually) or operate the complex visual interfaces of 3D printing software. The research developed a complete voice-controlled workflow enabling blind users to autonomously search for 3D models on the internet, select appropriate ones, configure printing parameters, and initiate the printing process. The system architecture evolved through two prototypes: an initial PC-based version proving technical feasibility, and an advanced smartphone-based version using a Raspberry Pi print server that abstracts hardware differences across 3D printers. The approach required solving multiple accessibility challenges beyond just voice control. For model search, the system combines a curated database of annotated models (with categories like technical, architectural, nature, persons, art) with retrieval from Thingiverse. Multi-part models requiring assembly are automatically filtered out since assembly typically requires vision. For print preparation, algorithms automatically handle placement on the build plate, determine optimal printing orientation to minimize support material, and calculate print duration. Users specify just two parameters: desired size (via maximum edge length) and quality level (low/medium/high). A key innovation is using water-soluble PVA filament for support structures, allowing blind users to simply soak the finished print in water rather than manually cutting away supports—a task that requires visual discrimination between model and support material.
Key findings
A requirements questionnaire completed by 52 people with visual impairments (average age 44) revealed strong demand: 84.62% found home 3D printing desirable, 90.38% wanted to print without sighted assistance, and 65.38% would invest 800 Euros in a suitable printer. Most participants had used fewer than 10 tactile 3D models in their lifetime, indicating limited current access despite educational value. The first user study with 7 participants (3 completely blind, 4 with minor residual vision) demonstrated technical feasibility. All participants successfully selected a model, initiated printing, removed the finished object from the build plate, and dissolved support structures in water. Average task completion time was 11.14 minutes (excluding print time). The main challenge was unreliable offline voice recognition requiring repeated commands. The second user study with 13 blind participants evaluated the smartphone-based interface for internet model retrieval. All participants successfully found and prepared a model for printing. Users averaged 2.46 search queries before selecting a model, spending 128 seconds on retrieval. Participants searched for items like Eiffel Tower, pyramid, switch, eagle owl, wheel rim, humerus, diesel engine, elephant, rat, dog, apple, and paraboloid. Users appreciated the autonomy but noted that text-to-speech was too slow and multilingual content caused confusion when model descriptions were in unfamiliar languages.
Relevance
This research demonstrates that complex physical fabrication processes can be made accessible through thoughtful system design. The key insight is that accessibility requires end-to-end consideration—not just an accessible search interface but also automated print preparation, hardware abstraction, and accessible post-processing (soluble supports). For practitioners, the work highlights the importance of autonomy as an accessibility goal. Many assistive technology solutions still assume sighted intermediaries; this research shows that with sufficient automation, blind users can independently operate equipment previously considered inaccessible. The annotated model database concept—where blind users share printing experiences and recommendations—could apply to other domains where user-generated accessibility metadata helps peers. The research also reveals practical challenges: voice recognition reliability, multilingual content in global repositories, and the difficulty of explaining spatial concepts like "maximum edge length" to users who cannot see reference objects. These findings inform the design of any voice-controlled system for visually impaired users.
Tags: blindness · visual impairment · 3D printing · tactile graphics · voice control · assistive technology · autonomy · speech recognition
Standards referenced: WHO visual impairment categories