From 8 to 23, the 3rd World Robot Conference (WRC) was held in Beijing. As a leading company in AI tactile perception, Tashan Technology showcased three key highlights at the event: the debut concept of "Robot Kindergarten," jointly developed with Turing Award winner Richard Sutton; the launch of two new tactile products—TS-FT multimodal sensor and TS-EST multimodal e-skin; and strategic partnerships signed with key collaborators including KanKan Intelligence, Xiechuang Data, Orbbec, and JD.com.
China Business Journal reporter observed that from technical concepts to hardware foundations and industrial ecosystems, Tashan Technology is building a complete closed loop centered on haptics.
Let robots grow through trial and error.
At the HeShan Tech exhibition zone, reporters saw a small robot that was the real eye-catcher: it wobbled as it walked. Rather than performing synchronized dance moves, it stumbled and paused like a toddler just learning to walk. This wasn't a sign of immature technology; it was a live demonstration of the "Robot Kindergarten" concept.
The "Robot Kindergarten" is a joint initiative by Tashan Technology and the team of Professor Richard Sutton, winner of the 2024 Turing Award and widely recognized as the "Father of Reinforcement Learning." It marks the first physical implementation of reinforcement learning theory in embodied intelligence. According to Tashan Technology, its core philosophy is that the real world is too complex for any pre-defined plans to cover all scenarios. Robots must grow through continuous interaction with their environment and trial-and-error—much like human toddlers—transitioning from an era of "passive data imitation" to one of "autonomous experience-driven evolution."
"Traditional imitation learning involves humans collecting data for robots to learn human movements," explained Dai Jingwei, Director of Ecosystem Operations at Tashan Technology. "In contrast, reinforcement learning enables robots to explore independently and refine their models through trial and error." The robot showcased on-site utilizes a reinforcement learning algorithm to continuously update model parameters in real time, iteratively improving its gait strategy during execution.
Making robots "durable" is the first step toward long-term trial-and-error learning. Powered by TS-EST's multi-modal e-skin for full-body tactile, thermal, and proximity sensing, the system detects when contact force exceeds safe thresholds. It then sends a negative signal to the decision model, enabling the robot to learn active avoidance of hazards. He Shan Tech calls this mechanism "giving robots pain perception."
The "Robot Kindergarten" is a five-year initiative scheduled to launch on 9/1. It is open to original equipment manufacturers, algorithm teams, research institutions, and scenario partners for co-construction. Professor Sutter attaches great importance to the project and has visited China multiple times in 2026 to directly participate in strategic discussions.
From fingertips to full-body
If "Robot Kindergarten" represents Tanshan Technology's exploration of future training paradigms, the two debut products showcase an upgrade in hardware capabilities—moving from "fingertip perception" to "full-body multi-dimensional perception."
The TS-FT multimodal sensor builds on existing 3D force, proximity, and position sensing by introducing temperature detection for the first time. It captures hot and cold contact changes within 3 seconds without requiring an external thermal probe, with a measurement range from -40°C to 120°C. This capability enables robots to perceive not just "force," but also "temperature," completing a critical loop in embodied intelligence operations.
The TS-EST multimodal e-skin is designed for full-body deployment on robots, integrating tactile, thermal, and proximity sensing. It achieves a force resolution of 0.01N, a temperature detection resolution of 0.1°C, and a proximity sensing range exceeding 1cm. This enables robots to perform safety assessments before physical contact—marking a qualitative leap in perception coverage from "a small patch at the fingertip" to "a full-body skin."
Additionally, Tashan Technology showcased the Emerald E10A Dynamic Tactile Sensing Chip—the world's first chip featuring a self-developed ASC module that converts tactile signals directly into pulse signals. It supports event-driven and sparse computing, offering microsecond-level time resolution and ultra-low power consumption.
The chip has completed tape-out. According to Dai Jingwei, the integrated sensing-and-control dexterous hand powered by Green Gem E10A is currently under joint development with partners. The goal is to transform the dexterous hand into a "standalone actuator" that achieves ultra-low latency response—similar to the human body's neural reflex arc—by closing the loop for perception and control locally.
Build a closed loop for the haptics industry
Recently, Hisan Technology has launched multiple partnerships across three key areas: non-robotic data collection, vision-tactile fusion, and channel expansion.
On 8/14, Tashan Tech and KanKan Intelligence achieved technical synergy by integrating visual perception and tactile interaction technologies to jointly develop a "Vision + Touch" multimodal perception system.
On 8 20, in collaboration with Createc Data, we launched a data acquisition solution to provide foundational support for capturing, processing, and training robotic operation data.
On 8 / 22, signed a strategic agreement with Orbbec. Orbbec's 3D vision provides spatial geometric data, while Tashan Technology's tactile sensing delivers physical interaction metrics such as contact force, temperature, and slip detection. Their integration is considered a key path to improving robotic manipulation success rates and safety.
On 8 23, Hisan Technology's tactile sensor products officially launched on JD.com, reaching enterprise procurement customers.
The logic behind these collaborations is clear: visual partners enhance our "seeing" capabilities, data partners strengthen our "collection" foundation, and channel partners streamline our "sales" pathways. Together, they are forming a closed-loop ecosystem centered on tactile perception.
"Previously, haptics were treated as an add-on, but this year companies want to leverage and maximize their potential," Dai Jingwei told reporters. This shift in perception has directly driven a surge in our haptic sensor shipments—first-half orders have already reached 2025 times the full-year volume of 4.
According to reporters, Hisan Technology has deployed intelligent shrimp-peeling equipment in Qianjiang, Hubei. Powered by its proprietary tactile perception technology, the system achieves a head-removal success rate of over 95% while fully preserving the shrimp roe.
Peeling shrimp is fundamentally a "twisting" motion—this perceptual capability can also be applied to industrial scenarios such as opening bottle caps, detecting cosmetic container seals, and inserting/removing optical modules. "We don't necessarily need to focus on humanoid robots," said the executive. "Using traditional robotic arms with sensors for industrial embodied AI offers lower costs and faster deployment."
The theme of WRC 2026 is "Human-Robot Symbiosis, Production-Needs Integration." From cutting-edge explorations in the "Robot Kindergarten" to hardware upgrades for two new products and a dense ecosystem deployment through numerous signings, Tashan Technology is demonstrating that tactile perception is moving from labs to production lines, becoming an essential underlying capability for robots to interact with the physical world.