The Need for a Prosthetic Knee Redesign

Researchers at Cleveland State University (CSU) are perfecting knee prosthetics to better mimic the human heal-to-toe walking gait. The design will take a page out of the electric car’s design to regenerate energy with each step, allowing the prosthetic to work longer. Additionally, the prosthetic is optimized using algorithms similar to the evolutionary factors that developed able-body legs.

 The problem with current knee prosthetic designs is that they work with a stiff walking gait. “If someone is walking with a stiff gait their knee is not flexing. When an able-body person is walking, the heel’s contact with the floor forces the knee to act as a shock absorber. Without that padding, it’s like walking on a peg leg putting stress on the hip. This can cause osteoarthritis and other types of back and bone problems,” explained Dr. Dan Simon, professor at CSU.

The prosthetic, on the other hand, will mimic the impedance and mechanical response of an able-body knee to external forces. The natural knee responses were recorded in the Parker Hannifin Human Motion and Control Lab. These responses were then mimicked in a digital design using MATLAB and Simulink before the researchers moved onto a real world prototype with motors.

The research spans three disciplines and three industry experts: Dr. Dan Simon for microprocessor designs and optimization, Dr. Hanz Richter for mechanical systems designs and controls, and Dr. Antonie van den Bogert who is a leading authority on biomechanics and gait analysis. Clearly, this isn’t a trivial design.

Merging Able-Body Knee Characteristics with Electric Car Technology

“The stiffness and inertia of a knee are like the shock absorbers and brakes of a car. The damping systems resist the force proportional to velocity. We are trying to control these characteristics to match an able body leg,” said Dr. Simon. “But we can also use this human braking system and regain energy much like an electric car.”

In an electric car, the braking system is very different from traditional motor vehicles. Instead of a brake pad the motor is engaged in reverse, acting as a generator. The energy to stop the car is then sent to recharge the batteries.

“We are doing the same with the prosthesis,” expressed Dr. Simon. “We don’t waste the heat from flexing the knee. Instead we engage the motor in reverse to act as a generator and store the energy in 650 Farad super capacitors. In this case, batteries are too slow to charge and discharge as the part of the stride where energy can be regenerated only lasts a couple hundred milliseconds.”

He added, “During a typical stride, the knee absorbs about 27 Joules of energy, which is wasted as heat. In our simulations, we can typically regenerate about 8 Joules of energy, which means that we have an efficiency of about 30%, and we can use the regenerated energy to provide power to the ankle.

Simulating an Able-Body Knee in a Prosthetic

Using MATLAB and Simulink the team was able to develop a baseline model of the system. “Simulink has a very nice graphical interface and the fact it is part of MATLAB makes it convenient and flexible when designing models,” explained Dr. Simon. “We coded all the low level control algorithm software in MATLAB and used Simulink for the higher level simulation and architecture. These control algorithms were based on the data we collected in the lab. We achieved close matching results.”

Simon added, “Simulation is important for any kind of engineering effort. It simplifies a lot of the effort to develop and optimize hardware, mechanical systems and control systems. A lot of these iterations and trade-offs are performed in simulation as opposed to building multiple expensive prototypes. Simulations are never 100% accurate, however, so you must still build a prototype for verification. Fortunately, a lot of the surprises are minimized in the simulation studies.”

To optimize the prosthetic knee’s design the team used an algorithm similar to natural selection, the driving force of evolution. “We see species evolve and adapt to better fit a suited environment one generation to the next," said Dr. Simon. "We can use a similar algorithm to better design the prosthetic. Evolutionary optimization software to produce a prosthesis to mimic an evolutionary derived system; it’s very poetic.”

In essence, the algorithm will use operators similar to those used in natural selection. These algorithms are commonly known as biogeography-based optimization, and particle swarm optimization. These optimization run will typically take 10,000 simulations running on 24 cores. Each simulation takes about one minute of CPU time. Therefore, the total compute time is about 6 hours using parallel computing on the 24 cores.

“The real testing starts when someone puts on the prosthesis,” said Dr. Simon. “Here you will discover modeling errors. You will need control algorithms that are robust enough to handle these variations in human use. But without simulation tools, we would never have got this far.”

Images courtesy of Cleveland State University.