Date of Award


Degree Type


Degree Name

Doctor of Philosophy in Electrical Engineering


Electrical, Computer, and Biomedical Engineering

First Advisor

Kunal Mankodiya

Second Advisor

Paolo Stegagno


This doctoral dissertation focused on the creation of wearables for gait which feature haptic feedback systems and can make use of Connected Environments (CE). The developed systems were created in a user-centric way with the intent of minimizing human effort for setup and configuration, using zero-configuration for network self-discovery and unattended organization.

CE empowers Internet Of Things (IoT) technologies to become non-stationary, enabling users to multitask, focusing on things other than the IoT Device. Since these technologies often are intended to have high availability and in some cases be always-on, discreet communication channels, such as the haptics of the vibration of a smartphone on mute, are becoming commonplace. Haptics can provide much needed feedback and are a useful design tool in crafting user experiences, especially in the context of accessibility systems. However, the physical space requirements and the design complexity of integration of these output components may preclude their inclusion in design considerations. While smartphones can be integrated into wearables to provide this feature, such as an exercise shoulder-band sometimes worn by runners which has a pocket for holding the smartphone, such designs require coupling additional hardware and restrict users both in access to their smartphone while also necessitating them to have their smartphone with them in order to function. Bluetooth paired end-points such as is common in smartwatch design, can weakly decouple the feedback as a subsystem or module, but the full networking advantages of wearable IoT (WIoT) remain unrealized as an intermediary host must provide WiFi access in many design configurations. These constraints can be alleviated by delineating the feedback module from the rest of the system with a purposeful haptic output system and by making it network accessible.

Likewise, the input system can also be isolated. This leaves the computational aspects to be freely relocated away from the user. This is consistent with the design paradigms often employed by household IoT devices or when part of a larger network of these devices or within a system, sometimes referred to as part of a “smarthome”. When these changes are applied to WIoT, some of the hardware may need to be duplicated and additional integration considerations must be evaluated. Similar to the advantages in system design that smarthomes provide through an uplink to remote systems for processing and storage, these Cloud servers have corollaries in healthcare CE. The same design principles of Cloud, Fog, and Mist computational paradigms spread out between the Edge (of which Mist is part of and Fog can be depending on architecture but may be kept separate for clarity) and remote computing network distances apply.

The experiment was to implement a secure infrastructure whereby an embedded system could eventually transmit arbitrary data to the Cloud. Since the device would not have a Cellular modem, WiFi was used. However, it was not reasonable to assume that the Cloud access would be fast enough for any response time needed for queuing gait or for near real-time healthcare alerts. An intermediary MQTT broker was then placed as a Fog host device and Zero-Configuration was built in. Initial testing of the system’s ability to be used with Fog-based AI was also evaluated.

After the creation and successful testing of a connected healthcare device for gait monitoring, various configurable aspects of the system were evaluated. Power constraints of the setup were also explored because as a wearable, charging tends to represent downtime where the system is not performing its desired task. Further, there is always a risk that the system will go off-line when the participants, or researchers, or if in a deployed setting, caregivers, are not aware of this state change. There may be a danger in this case that the activity may continue unmonitored. In situations where live monitoring is critical, this is an unacceptable scenario, so designing with power consumption in mind is important. Thus it was important to examine how the device’s screen and WiFi transmissions effected battery drain.

We then sought the reported opinions of amputees and whether they believed haptics would be helpful for them. The bilingual, English and Spanish, study consisted of 30 lower limb amputee prosthetic users. As our results indicated this to likely be beneficial and the sample population seemed willing to wear haptic feedback system with 80% reporting an interest in wearing prosthetics with haptic feedback and 30% reporting having already participated in therapy to improve balance, we started looking into what types of feedback can be provided. A slightly higher 83.3% of the total sample population reported believing that haptic feedback could help them.

How to best provide feedback to the user through haptics was explored. One of the more common techniques for adding haptics to a connected healthcare platform is to make use of the vibromotors present in many smartphones. Personal smartphones are often used because sample populations can typically provide their own and thus are likely already familiar and comfortable with that device.

Over 120 haptic patterns of a common haptic motor controller were evaluated for dissimilarity using 7503 zero-padding cross-correlations. Several mobile phone haptic patterns were also evaluated. The qualitative and quantitative evaluations led to the design of a new file specification. UHPP, a JSON structured human-readable format for describing the user’s haptic pattern preferences intended to make migration between haptic capable devices more portable.

Next presented was the evaluation of providing feedback to amputees with haptics based on force distribution of the prosthetic. The self-contained system, named HapticLink, connected force data from the bottom of the prosthetic to a band which featured haptic output. A series of evaluations were designed and two lower-limb amputees participated. The experiment consisted of a calibration step and several evaluations tests, in which the participants P1 and P2 scored accuracy results of 100% and 88.89% (averaged 94.44%), 66.67% and 91.67% (averaged 79.17%), and a perfect 100% and 100%, using the HapticLink. The initial tests and subsequent results established the utility of a simple, stand-alone force sensor and haptic motor feedback system to aid during single or double lower-limb amputee ambulation, using our system and its configuration.

Finally, alternative sensor systems are explored with a similar boot by developing a boot that can be interchangeable with the MQTT system previously presented. The evaluation of which, coupled with a design for haptic feedback system are left for future work.

Available for download on Friday, May 17, 2024