PANDORA

The PANDORA humanoid robot is a groundbreaking project developed at the Terrestrial Robotics Engineering and Controls (TREC) Lab at Virginia Tech, designed to explore the potential of additive manufacturing in humanoid robotics. Drawing inspiration from the THOR and ESCHER humanoid platforms previously developed by the TREC Lab, PANDORA represents a significant leap forward, integrating cutting-edge design and control methodologies to create a versatile, multi-purpose humanoid system.

At its core, PANDORA is more than just a robotic platform—it is an innovative tool for scientific research. The project goes beyond traditional engineering challenges, serving as an adaptable framework for testing in humanoid design, motion control, and real-world applications. Its modular and reconfigurable nature makes it a powerful asset for researchers aiming to test new theories of humanoid functionality on a platform that costs less than other robotic platforms and having an open-source code base with no black box.

A defining feature of PANDORA is its commitment to a hybrid design between additive and subtractive manufacturing. Unlike its predecessors, which relied heavily on traditional machining, PANDORA’s design leverages 3D printing to streamline prototyping, testing, and iteration. The machined pieces have been standardized with most being able to be used in multiple locations to reduce machining complexity. This approach offers unparalleled cost efficiencies, enabling rapid experimentation and design evolution, ultimately driving the project into new frontiers of innovation.

Beyond its cost-effectiveness, PANDORA is engineered for adaptability. Its modular structure and customizable control systems allow it to be tailored for a wide range of research applications, providing researchers the flexibility to modify both mechanical and software components to suit specific experimental conditions. This adaptability positions PANDORA as a dynamic tool for scientific exploration, bridging the gap between theoretical robotics research and real-world applications.

More than just a project, PANDORA reflects the pursuit of innovation and discovery in humanoid robotics. By leveraging additive manufacturing and modular design, it provides a versatile platform for research and development, enabling new approaches to robotic motion, control, and adaptability. As the project evolves, PANDORA continues to demonstrate the potential of cost-effective, customizable robotics, contributing to the broader advancement of the field. 

As of October 2024, I have completed my PhD and wrapped up my work on PANDORA. This project has been a transformative journey, marking my first deep dive into humanoid robotics. Over nearly seven years, I devoted my efforts to designing and developing a 3D-printed humanoid robot, culminating in the incredible milestone of seeing it take its first steps.

As the Head Mechanical Design Engineer, I gained invaluable experience in:

• Applying linear actuators for robotic movement

• Developing heuristics for designing structural components using additive manufacturing

• Constructing both the upper and lower body of a full-sized humanoid robot

• Collaborating with a dedicated team to turn PANDORA into a reality

• Manufacturing parts for both subtractive and additive manufacturing

• Applying Bio-Inspired designs to a humanoid robot

• And much more.

Working on PANDORA throughout my PhD was an extraordinary experience, blending technical challenges with creative problem-solving. While I accomplished a great deal, I only wish there had been more time to implement additional innovations and ideas that I had envisioned for the project.

During the Spring, my focus was on addressing persistent issues with PANDORA’s lower body and designing a bio-inspired upper body.

Lower Body Fixes

One of the primary challenges was encoder slippage at the knee joint. The original design relied on the knee joint pin remaining fixed in the lower knee to ensure accurate rotation tracking by the encoder. However, when the mounting piece failed to hold the pin securely, the encoder lost its position tracking of the knee.

To resolve this, I redesigned the encoder mounting system, making it a solid attachment point on the lower knee without relying on a connection to the pin to track the movement. Additionally, a new bracket was added to secure the encoder on the upper knee, allowing it to track movement reliably. This fix eliminated tracking errors caused by pin slippage.

Another issue was hip bracket failure due to unforeseen loading stresses along the layer lines of the 3D-printed part. This was mitigated by adjusting the print orientation, significantly improving the component’s structural integrity.

Bio-Inspired Upper Body Design

The highlight of this phase was the design and construction of PANDORA’s upper body, particularly the arms. Inspired by human biomechanics, the goal was to create more realistic joints using additive manufacturing, integrating tendon-like structures instead of traditional pin-based mechanisms.

Key features of the arm design:

• Motor Placement: 

  • To reduce weight on the elbow and wrist, all motors for the elbow and below were positioned in the upper arm, enabling increased payload capacity.

• Elbow Mechanism:

  • 1 degree of freedom (DOF)

  • Utilized two pivot points to maintain a constant cable length throughout its entire range of motion, ensuring consistent force transmission to the wrist.

• Wrist Mechanism:

  • 3 degrees of freedom (DOF)

  • Designed with an innovative twisting mechanism, mimicking the natural range of motion of a human wrist.

• Shoulder Joint:

  • Modeled after a ball-and-socket joint to replicate human shoulder movement.

  • Utilized tendon-like structures for stability, with a pulley system for controlled articulation.

This phase of development not only resolved critical lower-body issues but also introduced a new approach to humanoid joint design, blending biomechanics and additive manufacturing to push the boundaries of robotic motion.

During the Fall, the focus of PANDORA’s mechanical design shifted from developing new components to resolving emergent issues. This phase was largely "firefighting", addressing structural and mechanical challenges that arose during testing and operation.

Knee Joint Slippage

One of the primary issues was knee joint slippage. The existing design relied on a 3D-printed clamp that secured the knee joint by flexing around the shaft when tightened with a bolt. Over time, this approach led to minor variations in encoder readings, affecting the robot’s control system. To resolve this, we conducted a thorough analysis and implemented a redesigned clamping mechanism to improve stability and ensure accurate joint movement.

Hip Component Failures

A major flaw was identified in the hip’s structural integrity, specifically in how the actuator mounting point handled stress. The original print orientation caused the part to crack along the layer lines due to repeated loading. Since reorienting the print wasn’t possible due to bed size constraints, we implemented several reinforcement strategies:

• Thickening the part to improve durability.

• Adding epoxy reinforcement to increase strength.

• Integrating a thicker metal bar spanning the hip between both actuators, significantly enhancing structural integrity.

IMU Integration for Stability

A significant advancement was the installation of an internally mounted Inertial Measurement Unit (IMU) at the center of the hip. This upgrade allowed the controls team to utilize real-time IMU data to improve the robot's stability and balance—a critical step toward achieving locomotion.

Setbacks & Printing Challenges

This semester was not without setbacks. Repeated failures occurred while printing the redesigned hip, caused by:

• Filament feeding issues,

• Hot end failures,

• Z-axis motor failures in the 3D printer.

These technical problems delayed progress, with prints consistently failing around the tenth day of production.

Progress in Control Development

Despite these challenges, the controls team made substantial progress in fine-tuning PANDORA’s balance. Their advancements paved the way for the robot to take its first steps in the upcoming spring, marking a major milestone in the project.

This semester highlighted the resilience and problem-solving skills of the team, as we worked through mechanical challenges while making key improvements to PANDORA’s stability and structure.

The summer was a crucial period for the PANDORA project, with the primary focus on finalizing the arm design. One of the most significant challenges during this phase was developing a viable shoulder design.

Shoulder Redesign: Addressing Structural & Aesthetic Issues

The initial shoulder design, while functional, unintentionally gave PANDORA a "winged" appearance, and was shot down by the aesthetic committee. This feedback prompted a reassessment and creative redesign to achieve a more human-like structure.

To find a solution, I revisited a bio-inspired shoulder design originally proposed by one of my undergraduate team members. This concept provided a fresh perspective, and I worked on adapting it to:

• Align with PANDORA’s aesthetics

• Meet functional requirements

• Integrate seamlessly into the upper body chassis

The result was a significantly improved shoulder design, incorporating bio-inspired elements while maintaining structural integrity and mechanical efficiency.

Upper Body Completion & Printing Challenges

With the shoulder redesign finalized, we were able to complete the overall upper body design—a major milestone in PANDORA’s development. This allowed us to begin 3D printing the upper-body plates, which presented its own set of challenges.

Due to the size limitations of the CR-10 printer, the longest plate had to be printed horizontally rather than vertically. The entire print took 16 days to complete, creating a stressful situation where:

• Any failure after 7 days meant a complete restart,

• The print required constant monitoring to prevent mid-process errors.

Spring 2023 was a period of review and publication for the PANDORA project. The primary focus was on finalizing and publishing the first academic paper on PANDORA’s lower body, alongside continued progress in the control development.

Academic Paper on PANDORA’s Lower Body

One of the most demanding tasks this semester was the revisions and publication of the first academic paper detailing PANDORA’s lower body design and functionality. This required:

• Meticulous documentation of the development process

• Extensive validation and analysis of part design

• Structuring to effectively convey the scope and significance of our work.

This milestone not only showcased PANDORA’s technical innovations but also contributed to the engineering of humanoid robotics.

Technical Progress: IMU Integration & High-Level Control Development

Simultaneously, significant progress was made on the robot’s control system. The focus was on:

• Integrating an Inertial Measurement Unit (IMU) to improve PANDORA’s environmental awareness,

• Developing the high-level controller, a crucial step in refining motion planning and stability.

These advancements marked a major leap forward in PANDORA’s ability to interpret and interact with its surroundings, bringing it closer to achieving autonomous movement.

Maintenance & Minor Repairs

While much of the semester was dedicated to academic and control development, routine maintenance and minor repairs were also carried out. Though seemingly small, these tasks were vital in ensuring continuous functionality and minimizing disruptions in control system testing.

Conclusion

Spring 2023 was a dynamic period, balancing academic rigor with technical progress. On one hand, we worked to share our findings with the broader scientific community, while on the other, we continued refining PANDORA’s control systems, bringing the project closer to its long-term vision of having a walking humanoid.

During the Spring, our focus shifted toward addressing mechanical issues identified during the beginning of PANDORA’s control testing phase, this testing phase involved violent shaking of the parts along with motors going to the extremes as the control team began refining their code. As the controls team conducted real-world tests, they encountered various mechanical, electrical, and control-related challenges that required immediate fixes to maintain progress.

Mechanical Diagnostics & Rapid Issue Resolution

My primary responsibility was diagnosing and resolving mechanical problems that surfaced during testing. This involved:

• Conducting simulations to analyze failure points and identify root causes

• Implementing quick fixes to ensure minimal disruptions for the controls team

• Iteratively refining mechanical components based on real-time feedback

Conceptual Development of PANDORA V3

While troubleshooting PANDORA V2, I was also actively engaged in the early-stage development of PANDORA V3. This next iteration aimed to integrate:

• Key improvements identified during testing to large for quick replacements on V2

• More substantial structural and mechanical enhancements to ensure strength and reduce flex

• Lessons learned from PANDORA V2’s performance to optimize range of motion and other design choices

Conclusion

Despite encountering minor mechanical issues, PANDORA V2 proved to be structurally sound and reliable. The adjustments made during this phase not only enhanced its current performance but also validated the robustness of its design. These refinements provided critical insights that would directly inform the development of PANDORA V3.

Spring was a dual-focused period—on one hand, ensuring PANDORA V2 remained operational for testing, and on the other, laying the groundwork for PANDORA V3’s conceptual design. The iterative development process reinforced the project's adaptability and long-term vision, setting the stage for future advancements.

By mid-fall, the PANDORA project reached a major milestone: all components of the robot had been manufactured. After months dedicated to 3D printing and preparation of parts, we were ready to fully build the lower body of PANDORA. PANDORA V1 that had been used up until this point had a mix of parts from different changes and fixes along the way and PANDORA V2 would be a clean slate of all the same parts. 

Transition from PANDORA V1 to PANDORA V2

The assembly process began with a tear-down of V1 to get all electronics and wiring. PANDORA V1 was completely dismantled, allowing for careful inspection of the 3D printed parts to ensure there were no failures. To minimize downtime, a team of 2 disassembled and quickly reassembled PANDORA V2. With the coordinated efforts of two engineers, the full teardown and rebuild was completed in approximately 8 hours. This was an intensive but well-coordinated effort, ensuring a smooth transition from experimental development to refined execution.

Next Phase: High-Level Control Testing

With PANDORA V2 fully assembled, the team transitioned to the next phase of development; the control system.

Conclusion

The successful assembly of PANDORA V2 was a turning point in the project. It wasn’t just about constructing a robot; it was about bringing a refined, more standardized version of PANDORA to life. With the lower body complete, the project was one step closer to full-system integration and advanced testing.

Spring 2021 saw significant progress for the PANDORA project, as we embarked on the development of PANDORA V2 This version was a culmination of design insights and practical experience gained from the earlier version that looked to .

Key Innovations in PANDORA V2

The second iteration introduced several enhancements, including:

• Built-in circuit board mounts to streamline electronics integration

• Enhanced zeroing capabilities for encoders to improve precision and calibration

• A more modular and adaptable structure, allowing for swift reconfigurations to accommodate various experimental needs

These innovations were not only aimed at improving robotic functionality but also focused on optimizing the fabrication and assembly process, making PANDORA V2 a more versatile and efficient platform.

Conclusion

While the development of PANDORA V2 presented occasional challenges, the knowledge gained from designing PANDORA V1 proved invaluable. By applying these lessons learned, I ensured that this iteration was not just an updated version, but a more refined and effective design. Spring 2021 was more than just a phase of technical advancement; it was a demonstration of iterative learning and innovation. The development of PANDORA V2 highlighted the power of experience-driven design, paving the way for continued enhancements and breakthroughs in humanoid robotics.

Fall 2020 marked a defining moment for the PANDORA project, as the robot began to develop its own unique identity, diverging from its origins as THOR Mk. II. With a fully assembled lower body, PANDORA V1 became the focal point of development, particularly in troubleshooting and refining its functionality.

Advancements in Mechanical Design & Usability

My work during this phase was multifaceted, focusing on both structural improvements and quality-of-life enhancements, including:

• Creating and dispatching detailed drawings for machined parts

• Integrating sleeves and bearings to enable smoother motion

• Implementing design refinements to facilitate easier maintenance and handling

These improvements were critical in evolving PANDORA’s design, making it more reliable and operationally efficient.

Ball Screw Upgrade: Testing & Evaluation

Our team also explored upgrading PANDORA’s existing ball screws to improve performance. This effort involved:

• Sourcing and procuring newer ball screws,

• Custom machining for proper integration,

• Extensive testing and evaluation of the new components.

However, testing revealed that the new ball screws did not perform as well as the originals in key areas. As a result, we decided to revert to the original ball screws after some restoration, reaffirming the importance of thorough testing before implementing major hardware changes.

Conclusion

This period was marked by technical advancements, hands-on problem-solving, and practical learning. The challenges faced in design refinement and component evaluation reinforced the importance of rigorous testing in the development process. These lessons shaped PANDORA’s progression and prepared the team for future challenges and innovations.

Spring presented a new set of challenges for the PANDORA project, particularly when our initial full-leg prototype suffered damage during load testing due to a hard fall. The failure was traced back to the infill pattern used in printing, serving as a critical learning experience for our team.

Enhancing Print Robustness & Acquiring the CR-10 Printer

Undeterred by the setback, we focused on improving the durability of our 3D-printed components. A key step was the acquisition of a CR-10 printer for the TREC Lab, which provided new opportunities for experimentation. This upgrade allowed us to:

• Test various infill patterns and print orientations

• Explore new materials, expanding the scope of our research

• Improve print quality and strength, addressing previous mechanical weaknesses

Material Breakthrough: PLA+ as a Viable Option

Extensive testing with the CR-10 printer revealed the capabilities of PLA+ material:

• It withstood considerable loads

• It exhibited better elasticity comparable to ABS

• It provided a cost-effective alternative while maintaining structural integrity

These discoveries significantly influenced PANDORA’s future development, giving us a deeper understanding of material performance under load.

Optimizing Workflow & Transitioning to PLA+

Having the CR-10 printer also transformed our workflow:

• We could rapidly test design iterations at a much lower cost compared to MakerBot and Stratasys printers

• The shift to PLA+ simplified our printing process

• We were able to operate at lower temperatures, optimizing energy efficiency and print reliability

Conclusion

Spring was a period of setbacks, innovation, and discovery. The acquisition of the CR-10 printer, combined with methodical material testing, led to a breakthrough in print durability and the adoption of PLA+ as a primary material. These advancements not only strengthened PANDORA’s design but also accelerated development, bringing us closer to a more robust and efficient humanoid robot.

In Fall 2017, under the guidance of Dr. Leonessa, our team embarked on an innovative journey sparked by a fundamental question:
“Can we use additive manufacturing to build a humanoid robot?”

Recognizing the potential in this area, Dr. Leonessa put together a team of dedicated undergraduate students, including myself, to explore the practical applications of additive manufacturing in robotics. This initiative marked the beginning of what would later become PANDORA, a hybrid 3D-printed humanoid robot.

Laying the Foundation: Analyzing Existing Designs

At the project's infancy stage, PANDORA had yet to be named, but we had the foundational work of THOR and ESCHER as our starting point. Our initial focus involved:

• A comprehensive analysis of existing CAD models from TREC’s robotic lineup

• Evaluating the feasibility of integrating 3D-printed components into humanoid robotics

• Identifying opportunities where additive manufacturing could replace traditional machining

Expanding the Vision: A Fully 3D-Printed Robot

As the semester progressed, our team’s focus expanded beyond modifying existing designs. A segment of the group began exploring a more ambitious possibility:

• Instead of merely replacing select components, could we construct an entire humanoid robot through additive manufacturing?

• This idea gained momentum, leading to the design of full-scale, 3D-printed structural components.

A Pivotal Decision: Committing to Full Additive Manufacturing

By the end of the semester, after rigorous research and collaborative brainstorming, we reached a groundbreaking decision:

• We committed to building PANDORA entirely through 3D printing, venturing into uncharted territory in humanoid robotics

• Presented the lower-body prototypes to Dr. Leonessa, setting the stage for continued development

Conclusion

Fall 2017 was the inception of PANDORA, a project born from a simple yet transformative question. Our early research and decisions to create full parts through additive manufacturing, laying the groundwork for the future of humanoid robotics, setting the project on a path of continual innovation and discovery.