Previous Work

Otherlab - San Fancisco, CA

After leaving Google in October of 2013 I was recruited to lead a project at Otherlab building an robotic exo suit that would help soldiers run, jump, and do all sorts of insane things. We began by distilling down the requirements set forth by the government agency that was funding the project. Based on fundamental physics, many of the tasks they wished to perform were actually impossible, not hard, impossible. We then went through and searched for the performance metrics we could hit. Granted, these performance specs weren't achievable with any existing hardware available at that time. We decided to tackle running at high speed. This task requires a great deal of augmentation of a few key muscle groups. After studying a great deal of biomechanics texts and gait analysis research we noticed that calf muscles become activated early on as you ramp up speed then limit out. The knee acts somewhat similarly. After the ankle and knee are power limited, the hip flexors and extensors kick in to swing the leg even faster. We began to hypothesize what would happen if you didn't have these limits in your ankles and your knees. The following design is the result.

An initial prototype of a system that used our inflatable actuators was built in a very low fidelity way. We used a shin guard, some blocks of wood, and a manual control valve. With this we could see what it would feel like to have added torque in the ankle. We then built a higher resolution prototype that we could actually run around in.

After getting a feel for what the device would feel like I began sketching what a total knee/ankle system would look like. The design needed to remain ultra-light and provide half the power of a normal 100kg user running at 6mph. It would provide knee extension and ankle plantarflexion assistance at speeds of up to 12mph. This meant developing new actuators and new ultra-light valves to supply pressure accurately in the actuators.

A ton of work went into this 18 month project and we eventually delivered the first exoskeleton to provide metabolic benefit at a running speed over several different operators.

Meka Robotics - San Francisco, CA

From July '13 till Oct. '13 I developed research platforms for Meka Robotics during their acquisition by Google. My work there started with a forearm for an undersea robotic humanoid. The forearm consisted of 3 degrees of freedom, wrist roll, pitch and yaw.

As part of a contract job with ReNeu Robotics out of the University of Texas at Austin I designed and built a 14 degree of freedom torque controlled exoskeleton. The system has 5 actuators in the shoulder providing nearly the full range of motion of an able bodied user.

Ekso Bionics - Richmond, CA

From Sept. '11 to July '13 I worked for Ekso Bionics as a Senior Mechanical Design Engineer. The majority of my time at Ekso was spent working in the grants department developing next generation wearable robotic exoskeletons. These systems ranged from new actuated degrees of freedom for the medical flagship product, Ekso 1.0, load carrying systems to be implemented on the military system developed with Lockheed Martin, the HULC, and numerous passive exoskeletons that allowed users to carry heavy tools without injuring themselves.

The first project I completed at what was then Berkeley Bionics was funded by NSF STTR grant 0810782. The goal of this grant was to build a torque controllable ankle actuation system that could be implemented on an old version of the medical exoskeleton. Following several research papers from Hugh Herr and Greg Sawicki along with numerous biomechanics texts we opted for a torque range comparable to an average, able-bodied person. Torques were to be near 150Nm and a controllable bandwidth of about 7Hz. The final design, shown below was built and tested against our simulation results. The ankle actuators provided 140Nm of torque at a bandwidth of 6.5Hz which is more than enough to stand a person up on their toes very quickly. The actautor consists of a motor, brake, ballscrew, linkage, and a fiberglass spring in series. This series elastic architecture coupled with a sensor to read the deflection of the fiberglass beam allowed us to do high fidelity joint torque control. The sensing of the deflection of the composite beam is done by way of a living hinge linkage that is connected to the end of the spring and rotates an encoder at the base of the ankle.

Once I had assembled and verified the functionality of the ankle actuators, they were installed on a exoskeleton and control schemes were then developed. However, the other joints in the exoskeleton did not have a means for measuring torque at the joint level. To remedy this I developed several torque sensing mechanisms that were later installed in production units. Here are some examples of those parts. On the left is a knee pivot box that has been re-engineered to have a small cantilevered section that is instrumented with foil strain gauges in a full bridge to measure the bending moment in that beam which corresponds to the torque the actuator is outputting. In the small, plastic cover/pinch guard, we installed a small signal conditioning board that allowed us to calibrate the sensor before installing it into the unit. On the right there is another load cell design that measures the axial force in a ball screw. The lower right side of the part attaches to a linear rail to take any off-axis loads. You can see the small flexural parts just above the large round bosses. These allowed for misalignment of the ballscrew mounting. Again, a small signal conditioning board was installed to allow for calibration outside of the unit.

After the torque sensors were up and running and proven to perform as designed, I was moved over to work as technical lead for a hydraulically powered arm exoskeleton that had been designed by a previous engineer. This system consisted of two hydraulic actuators in the humerous area of the arm that allowed a user to carry heavy tools for long periods of time. Due to the lack of resources in the electrical engineering department, often times I found myself doing electrical debugging of the arm. This usually involved grabbing an oscilloscope and checking signals to and from the load cells, encoders, etc. To me, this can be one of the most rewarding jobs. There is nothing quite like having to figure out what is causing an issue in a complex system such as these and solving it.

During initial testing, we discovered that the torso that the user wore to hold the arm up was too large. My goal was to get the torso from 4 inches thick down to just 1/2 inch. To do this, I developed and printed out a torso that was then wrapped in carbon fiber for strength. The new design also implemented a quick release shoulder pin. This allowed the user to quickly mount and dismount the arm. Here are some photos of the process of prototyping that torso.

The final system ended up looking like this.

After about a month of work on the arm, I was tasked with creating a lightweight tool holding exoskeleton along with two other mechanical engineers. The main goal of this system was to implement a tool holder developed by Equipois called the zeroG arm. We developed, in roughly 2 weeks an entire exoskeleton that could lock and unlock hydraulic cylinders that went down to the users feet. This allowed for load transfer from the arm all the way to the ground.

Since lightweight is such an important feature with exoskeletons, especially passive ones, my focus in the year 2013 was to lighten them up as much as possible. For this, I used a similar process that we had used in making the torso for the arm to fabricate some lightweight legs for a new tool holding exo. With the use of composites, I was able to reduce the weight of the passive exoskeleton's legs from 4 pounds each to sub 1 pound. Here are the finished legs.

Bastian Solutions - St. Louis, MO

While a mechanical engineer at BMH Robotics I and a team of mechanical, electrical, and robotics engineers worked to provide the best robotic automation systems in the material handling industry. My role in the team included designing robotic end effectors for Fanuc, ABB, and Kuka industrial robots. These end of arm tools, EOAT's, utilized pneumatic, vacuum, electrical actuation, along with sensory systems that ranged from simple vacuum switches to 3D vision systems and all things in between. The purposes of these tools varied from simple vacuum tools to lift lightweight cardboard boxes to the more complex tools that were developed to lift roughly 300 different products all weighing 15 kg or more.

Other duties performed at BMH included generating detailed bills of materials, setting up suppliers for components, performing design reviews with customers to ensure their satisfaction with products, and working to support designs during the fabrication and assembly processes.

Above you can see our factory acceptance test of a system I designed that depalletized, cleaned, flattened, labeled, re-palletized, and shrink wrapped 50 pound bags of animal feed. My design responsibilities involved system layout, bag conveyor, the EOAT, cleaner, flattener, transfer conveyors, camera mounts, safety cages, and some of the pallet conveyor. Below, you can see a video of the system in action.

One other project at BMH was a pallet building machine for Green Line Armor. The system pressed injection molded bumpers onto the stringers, screwed down the lower cross boards, rejected any problematic pallets, flipped the pallets, screwed on upper deck boards, programmed the RFID tag, rejected any that needed rework, then stacked them. In this system, I designed the reject conveyor, pallet flipper, and pallet stacker. During bringup at our facility I was in charge of troubleshooting any technical issues as well as verify screw placement and throughput.

This EOAT was designed to palletize bundles of compacted aluminum scrap. While my job was to design the entire cell, the EOAT was the most interesting. Due to the fact that these "bricks' can come out anywhere between 5 and 7 inches long, we needed to measure each brick before it was put on a pallet in order to get the tightest pack. To do this I designed an auxiliary actuation system that was powered by Fanuc's electronics that drove a parallel gripper. One plate of the gripper was stationary while the other drove into the part. Once the part was contacted, we saw a current spike which we used as an indicator of contact. Using a string-pot we measured the length of the brick which allowed us to place the next brick right next to it. While this isn't the greatest way to do force control, it proved effective for dealing with this nasty material.

Harley Davidson Motor Company - Kansas City, MO

During an extended internship with Harley Davidson, I worked to improve the fabrication processes for fuel tanks, motorcycle frames and other sheet metal components. This often required use Pro/Engineer to design fixtures as well as modify sheet metal component models to improve the laser trimming programs. This internship gave me a great deal of experience in sheet metal forming processes and DFMEA in a large scale production facility.

Missouri University of Science & Technology - Rolla, MO

As an undergraduate student, I was appointed the position of teaching assistant for the course known as Intro to Manufacturing. As a TA I taught younger students proper use of machining and grinding machinery. This machinery included vertical and horizontal mills, lathes, drill presses, shapers, grinders, and band saws. The appointment of this TA'ing position helps to show my skills in machining operations as well as communication skills.