Evolutionary Touch Filter Chain Calibration

Abstract

Human machine interfaces (HMIs) are an essential part of our daily lifes. Generally, a shift from mechanical HMIs such as knobs to more electronics-based solutions can be observed. A well-known example are touch interfaces which may be found in a variety of products ranging from phones over cars to home appliances. Touch interfaces offer more usability but pose increased technical challenges which must be faced by the development team in order to meet customer demands. Additionally, legal obligations regarding their robustness must be fullfilled (even under noise) in order to allow the novel products to enter the market. This is often enabled by employing a dedicated signal filter chain which is to be set accordingly. The corresponding calibration process is often time intensive and error-prone if performed manually. This is for example the case for our industrial partner which is a major producer in the consumer electronics market. This work analyses if the pain of a manual calibration can be resolved. This task consists of probing new calibration(s) and measuring their performance as well as choosing a calibration. The former is automated by developing a corresponding testbed which can simulate various noise classes and can perform touch interactions on a real user interface using a six-axis robot. The novel calibrations are determined by a genetic algorithm (GA) which is a special optimizer out of the family of evolutionary algorithms. The optimization is the main focus of this study. Quantitatively the GA was analysed by benchmarking it successfully against several other optimizers as well as against a human expert. Furthermore, the search time of the GA could be reduced by improving the stopping criterion. The analysis is rounded up by a qualitative analysis which showed that the GA’s calibrations can withstand noise to a certain degree whilst still identifying touch events if and only if they occurred. Over all the study outlined that an automation of the manual process is possible on the scenarios considered.

A Forest Optimization Algorithm

Forest optimization algorithm (FOA) is another example for an evolutionary-based optimizer, which was introduced by Ghaemi et. al. [10]. In particular, it is a generic population-based meta-heuristic that is suitable for nonlinear continuous optimization problems. The algorithm is described in form of pseudo code in Algorithm 2. Its underlying concepts and details are described in this section.

The algorithm mimics the evolutionary reproduction strategy of trees in a forest. This strategy consists of different natural procedures of seed dispersal are used to spread the population of trees other the entire forest. The local environmental circumstances determine the survival of a subset of the population over time, which are representing the fittest individuals. In analogy to these procedures, a tree represents one solution of an optimization problem and the algorithm represents the evolutionary procedure of adding and selecting the best trees from the population.

FOA contains two types of search strategies, i.e., local and global search. Local search is performed by adding small variations to single variables of one tree, whereas global replaces them by randomly chosen new values. Please note in terms of local seeding (local search), a small random value r is added to $x_{select}$. In the presented FOA of Ghaemi et. al. [10], r is randomly chosen from the interval $[ -\Delta x, +\Delta x ]$. Additionally, the algorithm ensures to keep the affected variables in their ranges in the following way: In case of $x_{select}$ is exceeding its parameter range, $x_{select}$ is set to the upper limit of the range. In case of undercutting it, it is set to the lower limit, respectively.

Different hyperparameters are controlling the search behavior of the algorithm. These hyperparameters are briefly explained listed and explained:

• area_limit: The area_limit describes the maximum number of trees, that are allowed in the forest population.

• life_time: The lifetime parameter ensures that sub-optimal solutions, that are exceeding a certain age are removed.

• LSC: local seeding changes (LSC) control the number of trees created while the algorithm is performing a local search operation.

• $\Delta x$: During local search, a small value is added to the target variable, which is randomly drawn from the interval $[ -\Delta x, +\Delta x ]$.

• transfer_rate: transfer_rate determines how many former local trees are used for global search.

• GSC: During global search, global seeding changes (GSC) determine how wide-spread the global search is performed.

Please note: Similar as Ghaemi et al. [10] in their experiments, the initial number of trees needs to be equal to the area limit.

This section is concluded by a brief summary of the principle steps of the algorithm. In the beginning, an initialization phase creates a starting set of trees. Afterwards the algorithm starts its iteration process until a stop criterion is met. For each iteration the following steps are performed:

• Perform local seeding on 0-age trees.

• Control forest and form candidate population.

• Perform global seeding on candidate population.

• Update the best so far tree.

Once the stopping criterion is met, the best tree representing the fittest solution is returned.