Deformable and stretchable tactile sensors are significantly important in order to measure contact conditions over soft and unconventional 3D surfaces or stretching surfaces. Also, deformable tactile sensors which are sensitive towards deformation and stretch can help provide robots with humanlike tactile interaction capabilities, higher motion dexterity and better safety standards. However, lack of deformability and stretchability has been a well-known shortcoming of conventional tactile sensors. This is due to the presence of non-stretchable components such as numerous sensory elements and complicated wirings inside the sensing area.

We have taken advantage of Electrical Impedance Tomography, which is an inverse problem analysis method for estimating the resistance distribution of conductive materials, in order to develop a highly stretchable and deformable tactile distribution sensor which has no wiring in most of the sensing area. The developed sensor is able to detect not only pressure, but also other types of tactile stimuli such as deformation, stretch or heat. We have investigated the basic characteristics of the sensor including its sensitivity, accuracy and the effect of material hysteresis. Also, in order to achieve higher deformability and stretchability, we have developed our original conductive knitted fabric which is more deformable than ordinary conductive rubber and has less hysteresis. An original Pressure-sensitive Stretch-insensitive (PsSi) conductive structure has been developed as part of this research which enables stable measurement of pressure distribution independent from stretch conditions.

We have shown the abilities of this sensor in detecting rather complicated tactile interactions which involve skin deformation such as rubbing or pinching. Furthermore, we have demonstrated that the deformable tactile sensor can be extremely useful for implementation over complex 3D surfaces such as the robot face. Finally, we have overcome a major challenge in field of tactile sensors by presenting a successful implementation of the stretchable tactile distribution sensor over movable joints. We have been able to detect external tactile stimuli over the joint surface at different joint angles and have estimated the joint angle value by using only the tactile distribution patterns from the tactile sensor. We have argued that the ability of the tactile distribution sensor to detect joint movements can be used to provide the robot with a somesthetic feedback which can lead to a sense of body image.

Finally, applications of the sensor particularly in wearable tactile systems and intelligent furniture are addressed. We have realized stable sensing of externally applied pressure over dynamically deforming and stretching surfaces such as the area around human elbow joint and the hip area, by minimizing the effect of stretch and pressure caused by body movements. This has not been possible using any of the conventional tactile distribution sensors.

This thesis is organized in 7 chapters as follows:

Chapter 1 (Introduction) presents the main problems with conventional tactile sensing technologies. The main characteristics of the human sense of touch and the centric role it plays in our everyday tasks are explained. We have argued that the dominant approach toward tactile distribution sensing in the past decades, has been the sensory array approach which houses numerous sensory element in a base material and samples them independently. We have argued about the limitations of this approach with regards to stretchability and deformability, which in turn are rooted in the complex wiring inside the sensor sheet. The main objective of this thesis, which is to introduce high stretchability and deformability to the tactile distribution sensors through elimination of wiring inside the sensor area, is presented and the contributions of this research are clarified.

Chapter 2 (Stretchable and Deformable Tactile Sensors) defines the concept of stretchability and deformability with regards to other commonly used terms such as soft and flexible and argues the necessity of stretchable and deformable tactile distribution sensors. It argues that, measuring the conditions and dynamics of contact on a target surface, requires a tactile sensor which is more conformable than the surface itself. Some previous works on stretchable tactile distribution sensors are presented and the novel approach of this research towards realizing stretchability and deformability through Inverse Problem Analysis is explained. Prospects of stretchable tactile distribution sensors in a number of different fields, such as medical and rehabilitation, robotics and wearable sensing devices are also discussed in this chapter.

Chapter 3 (Tactile Sensors Based on Inverse Problem Analysis) presents the main concept of Electrical Impedance Tomography(EIT), which is an Inverse Problem technique for estimating the internal resistance distribution of a conductive medium using measurements on the boundary. Mathematical framework of EIT, including the forward and inverse problems is presented and a typical EIT system is introduced. The implementation of EIT to tactile sensing and the characteristics of an EIT-based tactile sensor have been discussed. Measures for resolution and sensitivity of the sensor, such as Minimum Detectable Area, Minimum Detectable Resistance Contrast, Sensitivity and Selectivity distribution and distinguishability have been introduced and the effect of different modeling parameters on the resolution of the sensor is addressed through simulation. Furthermore, this chapter includes a brief section on the expansion of sensor coverage area and increasing the number of electrodes. The results of this section are independent of the material used as the conductive medium in the sensor.

Chapter 4 (Fabrication of EIT-based Tactile Sensors) presents the step-by-step guide to fabricating the proposed tactile distribution sensor, covering the data acquisition device, developed software and suitable conductive materials. Detailed characteristics of several conductive materials such as many types of conductive rubber, films and fabrics have been presented and their problems have been categorized into (1) high hysteresis and (2) indistinguishability of pressure and stretch. The first problem has been addressed by developing an original conductive fabric with very low hysteresis, and experiments demonstrating this superiority are performed. We have argued that although solving the second problem could have two major solutions, it would ultimately require a conductive medium which is only sensitive to one particular tactile stimulus (pressure or stretch). The design and development process of a fabric-based PsSi (Pressure-sensitive Stretch-insensitive) conductive structure has been presented and the response of the sensor towards pressure and stretch stimuli are demonstrated through experiments. Finally, the issue of electrode connection to the medium is addressed and our original stretchable fabric-based electrodes have been introduced. As a result, this chapter elaborates on the detailed process of design and development of a fully fabric-based tactile distribution sensor which is highly insensitive towards stretch and shows a very small hysteresis effect towards pressure.

Chapter 5 (Tactile Sensing Scenarios) categorizes the different surface types in which tactile sensing is required into (1) Flat rigid surfaces, (2) Complexly curved rigid surfaces, (3) Passively deforming surfaces and (4) Actively deforming surfaces. In each of the cases, we have presented a tailor-made sensing solution focusing specifically of the suitable conductive medium for the sensor. In case of actively deforming surfaces, the two solutions for pressure-stretch indistinugishability presented in chap.4 are implemented and the results of experiments are shown. This chapter also presents a solution for minimizing the effect of self-movements in an actively deforming surface on the sensor medium, which will enable us to detect only external pressure over the surface.

Chapter 6 (Applications and Results) demonstrated the applications of the fabric-based sensor in areas such as wearable sensors and intelligent furniture. We have implemented fabric-based PsSi sensors over highly deformable and stretchable areas such as human hip and elbow. Such areas are selected due to the high deformability and stretchability of the surfaces which has made it impossible for conventional tactile distribution sensors to cope with. Also, the implementation of the tactile sensor over the deformable surface of a chair is presented while demonstrating the possibilities of using the tactile distribution results as a control signal in intelligent furniture.

Chapter 7 (Conclusions and Prospects) presents a brief summary of the thesis and the full conclusions of this work. The contributions and originality of the work are addressed in detail, and the major characteristics of the developed tactile sensor are reiterated. This chapter discusses the general problem of low special resolution in EIT measurements and provides a number of solutions into improving the resolution in EIT-based tactile distribution sensors. Lessons learned from each application and the practicalities of the developed sensor are discussed. A brief section of this chapter discusses the commercialization prospects of the sensor including the production cost and potential roadblocks. Finally, some of the future directions and goals for this research are mentioned including the possibility for Stretch-sensitive Pressure-insensitive (SsPi) sensors, applications of the sensor in user interfaces and virtual reality environments and implementation of the sensor over deformable robots.

本論文は,「Development of a Highly Stretchable and Deformable Fabric-based Tactile Distribution Sensor」と題し,これまで実現困難とされてきた伸縮・変形可能な触覚分布センサを,従来のアレイ型とは本質的に異なる,センシング領域内に配線等を持たない新たな構造および計測方式により実現し,さらに構造や材料の工夫により伸縮の影響なく圧力のみの検出を可能とし,これらの有用性を基礎実験や応用例により実証した研究をまとめたものであり,7章からなる.ロボットの触覚センサは20年以上の研究の歴史があるが,特に最近,人間支援ロボットにおける安全動作や全身接触動作等のために,全身を触覚で覆うことが重視されている.これを完全に達成するためには,関節部などの可動,変形部や柔軟な表面を覆うことが必要となるが,従来の分布触覚センサは大きな伸縮・変形を許容しないため,きわめて困難な課題であった.以下の各章では,逆問題解析に基づく触覚センシング方式の提案,センサ特徴および拡張性の検証,センサの開発及び基礎実験による評価,伸縮・変形可能な実装面への適応,アプリケーションにおける実用性の実証について記述し,学術的考察を加えている.

第1章"Introduction"では,触覚センシング研究の背景について述べ,伸縮・変形の観点から人間の触覚と既存触覚技術との比較を行い,解決すべき課題を呈示している.また,大伸縮・変形に対応しながら安定に触覚刺激を計測することを目的にし,この目的の学術的な貢献について論じている.

第2章"Stretchable and Deformable Tactile Sensors"では,本研究で扱っている「伸縮」および「変形可能」の用語を定義し,特にソフト・フレキシブルとの相違を明確化した上で,触覚センシングにおける伸縮・変形可能なセンサの重要性について論じている.また,伸縮性触覚分布センサを実現するための既存手法を紹介し,提案手法との関連性と相違について記述し,逆問題に基づく提案手法のオリジナリティを明確化している.また,提案手法による触覚センサが関連分野にもたらす影響やこれからの展望について議論している.

第3章"Tactile Sensors Based on Inverse Problem Analysis"では,逆問題の一種であるElectrical Impedance Tomography (EIT)の基本概念や順問題・逆問題の数理モデルを紹介し,EITに基づく触覚分布センサの原理について記述している.具体的には,触覚刺激で抵抗分布が変化する導電シートを用意し,その辺縁のみに設置した複数電極により電流注入と電位計測を行い,このデータからEIT手法により間接的に導電シート内部の抵抗分布を推定し,触覚刺激を推定する.原理説明に加えて,センサのサンプリング手法,空間分解能,感度,各種パラメータの影響など導電材料の特性に無関係な特徴をシミュレーションで評価している.

第4章ではセンサの具体的実現法を詳しく説明している.計測用の回路とアルゴリズム,各種導電材料の特徴を記述した後,導電性材料のヒステリシス問題や伸縮・変形の影響を解消できる導電性材料・構造の必要性について論じ,伸縮・変形に不感な感圧導電性構造(PsSi: Pressure-sensitive Stretch-insensitive)を布材料で実現する方法を提案し,製作したPsSi構造の圧力や伸縮に対する応答を記述している.さらに,独自に開発した布製の伸縮性配線も統合し,電極も含めて布材料のみで実現されるセンサモジュールを提示している.

第5章"Tactile Sensing Scenarios"では,提案したセンサの適用状況として,(1)平らで固い面,(2)複雑な形状をした固い面,(3)変形可能な柔らかい面,(4)動的に変形する柔らかい面,の4種類を挙げ,各状況に対応するセンシング方法を提供した.特に,従来計測が難しいとされていた(3)と(4)において,PsSiセンサを利用して安定に圧力分布を計測できることを示している.

第6章"Application and Results"では,開発したセンサを人間の肘や臀部に実装し,人体の動きに影響を受けることなく外部からかかる圧力分布を計測できることを示した.これにより,既存の分布触覚センサで不可能とされてきた伸縮・変形可能な実装面における安定な圧力分布計測が可能であることを実証し,ウェアラブルセンサへの応用可能性を提示した.また,インテリジェント家具にむけた応用事例として,椅子に実装したセンサでの実験結果に基づき,触覚パターンを制御信号として利用する可能性を議論している.

第7章"Conclusions and Prospects"では,本研究を総括し,その寄与は,伸縮性触覚分布センサの原理の提案と基本特性の検証,安定な触覚センシングに必要な導電性材料・構造の開発,各適用状況に対応したセンシング方法の提示を行い,ウェアラブルセンサやインテリジェント家具等の応用事例の実験によってその実用可能性を示したことであると結論づけている.

以上,これを要するに,本論文は,(1)EIT原理に基づく伸縮性触覚分布センサの実現,(2)シミュレーションによるセンサ基本特徴の検証,(3)伸縮・変形の影響を受けない触覚センシングの方法論,(4)各適用状況に応じたセンシング方法が明らかにし,(5)具体的な応用実験により提案センサの有効性を示している.これにより,従来不可能であった伸縮・変形可能な面における触覚分布センサの実装が可能になり,伸縮・変形の影響を受けない安定な触覚センシングが実現された.この成果はロボット学における一つの懸案を独創的な方法で解決したものであり,マンマシンインターフェース,人間工学,医療・リハビリテーション等への新たな応用可能性においても有意義と認められる.

以上の理由から,本論文は知能機械情報学上貢献するところ大である.よって本論文は博士(情報理工学)の学位請求論文として合格と認められる.