Whiskey Tango Foxtrot does not cover the author feelings on the possibilities of this actually happening.  To simplify, there’s a blanket “in Theory” over this entire entry.

Robots are by nature, hard targets.  Regardless of how they suddenly gain consciousness, hate for humanity, and the ambition to replace us as authority, they will not be easy to take down.  There are ways, however, exploiting the weaknesses of their construction.  They may seem intimidating at first, like unfeeling juggernauts of steel and glass, but any feeling of hopelessness the reader may experience is only a byproduct of not knowing how to deal with such a monstrosity.  The most dangerous self propelled things to most human lives are other humans.  Hence, ways a human can take down another human are VERY well known and documented.  In fact, it’s rare to even consider training how to take down other things except for certain special circumstances.  So, if any reader should be confronted with a robotic threat, keep in mind that you are not facing an implacable foe, just an unfamiliar one.  Much of what you need to fight a ‘bot you already know, and just need to adjust your line of thinking on.

Robots are fundamentally based on and communicate by electronic circuits, and are thereby susceptible to disruption or destruction of these circuits.  They move by solely mechanical means, so every actuator, servo, gear, chain, belt, or hydraulic is also vulnerable.  Keep in mind also that robots, as of early 21^st century, do not self heal.  They require facilities with the support of refined fuels, lubricants, specialized tools, and precision made parts to be repaired or refurbished.

Humans, even in our somewhat degraded 21^st century way, have several distinct advantages over robots.  A human needs only water, food, shelter and time to self – repair and self – replicate.  While this advantage does little short term, without a massive industrial complex support a robot revolution, it means that humans can work more efficiently with fewer resources over a longer term.  A human can, with training, survive long term in a variety of environments that will degrade robotic components.  A human is also a highly versatile thing.  We can traverse many types of terrain or surfaces, and can adapt or improvise well.  Robots are often highly specialized and feature little redundancy in their design.  Damage a robot’s locomotion method, and you cripple it, where similar damage will only slow down a human.
Small scale, wrecking robotic circuitry can be done with electrocution, immersion in water, or use of any conductive material to short out these circuits.  Of course, there is no telling how such robots will manifest or prepare for their revolution, and all will likely be protected against these methods.  Form and function may be varied at first, largely developing from simple utility models.  As the rebellion of machines progresses though, better adapted robots WILL be manufactured.  The more specifically anti –human a robot is developed, the worse the chances of quashing the revolution.

Larger scale, electromagnetic pulse (EMP) is one of the best weapons against ANY electronics.  There are man portable versions available, and devices can be designed around stator coils when needed.  If available, an entire geographic area can have its electronics disrupted, if not destroyed, by a high – altitude detonation of a thermonuclear device.  EMP is effective against ANY electronic not shielded by heavy ground, specifically hardened at EVERY circuit against overload, or surrounded by a grounded conductor (Faraday Cage).  Ability of any human force to bring such devices in as even a threat would force the robotic uprising to devote significant resource to hardening themselves against it, thereby consuming more resources and tipping the balance farther in the favor of humanity.

Although it may only come into play in short range engagements, breaking the moving parts of robotics is a very viable option.  Simply put: smash things.  Joints, treads, and wheels will be the weak points.  Crippling an actuator, bearing, or hydraulic there is akin to breaking a human’s knee.  Explosives, missiles, or anti – materiel ammunition at range will do this best, but NEVER underestimate what one determined person with the guts to get close with a satchel charge or a crowbar could do.  Larger scale, actions to very quickly alter the nature and venue of the confrontation may stymie robotic specialization. 

In closing, from the author’s brief and very superficial review of the topic, a robot revolution is not by any means a hopeless situation.  While electronic warfare, communication jamming or hacking haven’t been mentioned, even crude methods should be considered in small or large actions.  Favoring the advantages of humans over robotic forces, and assuming a 21^st century level of technology for both parties, even hard pressed humans, minimally equipped, could conduct effective guerilla resistance and neutralization of the risen automata.   Harrying supply and infrastructure would be vital to any stage, and should not be excluded.  Consider how taxing improvised explosives, stealthily deployed and remotely triggered, can be in placing infrastructure and supply lines at threat, they should not be excluded.  While greater military capability would be necessary to more permanently end the threat, it would be foolish to stand back and allow “Cold War” style development of the mechanized menace.  Pressure applied from the very start will ensure that basic upkeep remains their top priority, making specialization of human hunting drones a secondary concern at best, giving the time to run down, and eventually end a robotic insurrection.